DETAILED DESCRIPTION

[0031] Referring now to the drawings, FIG. 1 represents a block diagram of an embodiment of a mosquito noise reduction apparatus 50 in accordance with the invention.

[0032] MNR apparatus 50 receives four (4) main system inputs. Image signal luminance Y and chrominance Cu/Cv components are applied at inputs 100 and 101 u/v respectively. Coding Parameters at input 102 might represent, for example, an average coding bit rate. In the preferred implementation, this input is simply controlled by an end-user in a heuristic manner. Also, the end-user controlled Signal Mode signal 103 represents the thresholding values pre-determined for an image signal type such as DVD, DSS, DV-25 signal etc.

[0033] MNR apparatus 50 comprises five (5) main blocks: image Region Classifier (RC) 104 , LUminance component REgion-Based Noise Power Estimator (LU-REBNE) 106 , LUminance component LOcal SEgmentation-based Adaptive Noise Reducer (LU-LOSEGANR) 108 , CHrominance component Local Noise power Estimator (CH-LONE) 112 and CHrominance component LOcal SEGgmentation-based Adaptive Noise Reduction (CH-LOSEGANR) 115 . It is important to note that, for simplicity, FIG. 1 illustrates only one CH-LONE 112 and its associative CH-LOSEGANR 115 for both chrominance components Cu and Cv. Persons of ordinary skill in the art will understand that such components may be implemented in a time sharing manner or in parallel as is well known in the art.

[0034] Image Region Classifier (RC) 104 described in detail below with reference to FIG. 2 receives three (3) signals, namely: the decoded luminance Y 100 , an interpolated chrominance Cu/Cv 116 u/v and the signal mode 103 to generate a region map 105 . Image Region Classifier 104 is responsive to the luminance component Y 100 of the decoded image signal for analyzing each luminance pixel value thereof in accordance with a corresponding luminance pixel spatial context in a same frame of said image signal. RC 104 classifies the luminance pixel in a selected one of a plurality of predetermined image region classes associated with distinct image region spatial characteristics and generates a corresponding selected region class indicative signal (Region Map 105 ). Conveniently, the predetermined image region classes or region map allows the classification of a current pixel as belonging to an edge (E), a flat region (F), a near flat region (NF), a near edge region (NE) or a finally textured (T) region, as distinct image region spatial characteristics.

[0035] Region map signal 105 , luminance signal Y 100 and Coding Parameters 102 are applied as main inputs to the Luminance component REgion-Based Noise power Estimator (LU-REBNE) 106 . Two (2) secondary signals 110 and 111 that represent data on the segmented local window generated by the LU-LOSEGANR 108 are also applied to LU-REBNE 106 . LU-REBNE is a shape-adaptive luminance noise power estimator that is responsive to the luminance component Y 100 and the selected region class indicative signal (Region Map 105 ) for estimating statistical characteristics of the luminance pixel by using local window segmentation data associated with the luminance pixel, to generate a corresponding luminance noise power statistical characteristics indicative signal.

[0036] LU-REBNE 106 described further below with reference to FIG. 3 yields an estimated luminance noise local standard deviation signal 107 in the decoded luminance component. The noise local standard deviation is required further for a MMSE noise reduction.

[0037] Noise local standard deviation signal 107 and noisy luminance component Y 100 input to LU-LOSEGANR 108 which yields, in turn, a filtered Y luminance signal 109 and the two signals 110 and 111 containing data on the segmented local window characteristics. LU-LOSEGANR is a shape-adaptive luminance noise reducer for filtering the luminance component Y 100 according to the luminance noise power statistical characteristics indicative signal (noise local standard deviation signal 107 ). LU-LOSEGANR 108 is described further below with reference to FIG. 4 .

[0038] Chrominance Cu/Cv signals 101 u/v , Coding Parameters signal 102 and the segmented local window data signals 110 and 111 are applied to CHrominance component LOcal Noise power Estimator (CH-LONE) 112 . CH-LONE 112 provides an estimated chrominance noise local standard deviation signal 113 in the chrominance component, required for a chrominance MMSE noise reduction as is described further below with reference to FIG. 5 .

[0039] Finally, chrominance noise local standard deviation signal 113 and noisy chrominance Cu/Cv signals 101 u /v are input to CH-LOSEGANR 115 . CH-LOSEGANR 115 yields, in turn, interpolated chrominance components signals 116 u/v optionally required in the RC block 104 , and filtered Cu/Cv chrominance signals 114 u/v . CH-LOSEGANR 115 is described further below with reference to FIG. 6 .

[0040] As is understood by persons of ordinary skill in the art, appropriate delays for signal synchronization required by the various operations of MNR apparatus 50 are not illustrated. Implementation of such delays is well known in the art.

[0041] Referring now to FIG. 2 , there is illustrated in block diagram Region. Classifier (RC) 104 in accordance with the invention.

[0042] A decoded noisy luminance signal Y 100 is applied to the region classifier RC generally designated by 104 . Firstly, for a reliable classification, the noisy signal Y is filtered by a diamond shape 2D diagonal low pass filter L 1 ( 201 ) in order to reduce high frequency noise component. The filter impulse response is given by the following equation: 1 $\begin{array}{cc}{L}_1\left(i,j\right)=\left[\begin{array}{ccc}0& 1& 0\\ 1& 4& 1\\ 0& 1& 0\end{array}\right]/8& \left(1\right)\end{array}$

[0043] in which the couple (i, j) represents the current coordinates (line, column) of the central and considered pixel. The filter output 202 is sent to four (4) Sobel edge masks 203 , 204 , 205 and 206 designated respectively for 0°, 90°, 45° and 135°. Their respective impulse responses are: 2 $\begin{array}{cc}{\mathrm{Sobel}}_0\left(i,j\right)=\left[\begin{array}{ccc}1& 0& -1\\ 2& 0& -2\\ 1& 0& -1\end{array}\right]& \left(2a\right)\\ {\mathrm{Sobel}}_{90}\left(i,j\right)=\left[\begin{array}{ccc}1& 2& 1\\ 0& 0& 0\\ -1& -2& -1\end{array}\right]& \left(2b\right)\\ {\mathrm{Sobel}}_{45}\left(i,j\right)=\left[\begin{array}{ccc}0& -1& -2\\ 1& 0& -1\\ 2& 1& 0\end{array}\right]& \left(2c\right)\\ {\mathrm{Sobel}}_{135}\left(i,j\right)=\left[\begin{array}{ccc}2& 1& 0\\ 1& 0& -1\\ 0& -1& -2\end{array}\right].& \left(2d\right)\end{array}$

[0044] Each of the Sobel masks 203 , 204 , 205 , and 206 has a respective output, 207 , 208 , 209 and 210 , to a respective absolute value detector 211 , 213 , 215 and 217 . The respective outputs 212 , 214 , 216 and 218 of the detectors 211 , 213 , 215 and 217 are now utilized for two different purposes: strong edge detection and flat region detection.

[0045] For edge detection, the four (4) absolute value detector outputs 212 , 214 , 216 and 218 are applied respectively to their associated thresholding (comparison) operators 220 , 222 , 224 and 226 . The thresholding output is equal to 1 if its corresponding input is greater than or equal to a threshold value; otherwise, it will be 0. The pre-determined threshold values at 272 are given by a Look-Up Table (LUT) 273 that is controlled in turn by Signal Mode signal 103 . The four comparison operator outputs 229 , 230 , 231 and 232 are applied together to an OR gate 237 whose output 239 represents a preliminary detection for strong edges in a given image. This detection is far from perfect; the detected edge can be broken or composed of isolated points. To partly remedy the situation, the preliminary detection binary signal 239 is submitted to two (2) non-linear operations in cascade. The first one, Add Only Consolidation (AOC) 241 , is defined as follows: Consider a local window centered on the current pixel. If a count of the “1” number in the window is greater than or equal to a threshold, then the operator output is “1”; otherwise, the output remains unchanged. In the preferred implementation, the window dimension is 3×3 and the threshold value, at 240 , is set to be 4. The AOC operator can be described by the following:

[0046] Let in(i, j) and out(i, j) denote respectively the input and the output of the operator at the coordinates (i, j) of the current pixel. Let W is the local window domain. The operator output is given by: 3 $\begin{array}{cc}\mathrm{out}\left(i,j\right)=\{\begin{array}{cc}1,& \mathrm{if}\sum _{\left(n,m\right)\in W}^{}\mathrm{in}\left(i-n,j-m\right)\ge \mathrm{Threshold}\\ \mathrm{in}\left(i,j\right),& \mathrm{otherwise}.\end{array}& \left(3\right)\end{array}$

[0047] The second operator 248 , Remove Only Consolidation ROC, is in turn given by: 4 $\begin{array}{cc}\mathrm{out}\left(i,j\right)=\{\begin{array}{cc}0,& \mathrm{if}\sum _{\left(n,m\right)\in W}^{}\mathrm{in}\left(i-n,j-m\right)\le \mathrm{Threshold}\\ \mathrm{in}\left(i,j\right),& \mathrm{otherwise}.\end{array}& \left(4\right)\end{array}$

[0048] In the above equation, in(i, j) and out(i, j) are respectively again the input and the output of the considered operator and W is the local window domain. In other words, if the count of “1” numbers in the window is smaller than or equal to a threshold, then the operator output is “0”; otherwise, the output remains unchanged. In the preferred embodiment, the window dimension is 3×3 and the ROC threshold at 245 is equal to 2. The ROC output signal 251 represents now the detected edge map.

[0049] In order to determine a Near Edge (NE) region, the detected edge map signal 251 is block-based expanded by a binary operator Block-based Add Only Consolidation (BAOC) 253 . In the preferred embodiment, the block dimensions are 4 lines by 8 columns. There are a few reasons for these chosen dimensions: first, in some CODECs for recording mediums such as DV-25, DV-50, the block dimension can be 4×8 and in the popular MPEG-2, the compression blocks can be frame-based 8×8 (i.e. in a given field, the dimension of a block is 4×8); second, 4×8, which is a sub-block of 8×8, has been experimentally proved to be a compromise between over-correction and picture naturalness preservation. BAOC 253 is described as follows. In a given block, if the number of edge pixels, represented by a number of “1”, is greater than or equal to a threshold, (e.g. 3) at 259 , then all pixels in the block become “1”; otherwise, the block remains unchanged. Let B be the considered block domain. The descriptive equation is given by the following Equation (5): 5 $\begin{array}{cc}\forall \left(i,j\right)\in B,\mathrm{out}\left(i,j\right)=\{\begin{array}{cc}1,& \mathrm{if}\sum _{\left(i,j\right)\in B}^{}\mathrm{in}\left(i,j\right)\ge \mathrm{Threshold}\\ \mathrm{in}\left(i,j\right),& \mathrm{otherwise}.\end{array}& \left(5\right)\end{array}$

[0050] BAOC output 258 is then applied to an AND gate 260 together with the negation of binary edge signal 251 . Black dots at AND gate inputs denote negation of the considered input in FIG. 2 . AND gate output 268 from AND gate 260 represents the detected NE region map signal.

[0051] For a flat region detection, the four (4) absolute value detector outputs 212 , 214 , 216 and 218 are applied respectively to four (4) other associated thresholding (comparison) operators 221 , 223 , 225 and 227 . The thresholding output is equal to 1 if its corresponding input is smaller than a threshold value; otherwise, it will be 0. The pre-determined threshold values at 228 are given also by LUT 273 that is controlled in turn by Signal Mode signal 103 . The four comparison operator outputs 233 , 234 , 235 and 236 are applied to an AND gate 238 whose output 242 represents a preliminary detection for flat regions in a given image. This flat region detection can be composed again of isolated points or isolated holes. To partly remedy the situation, preliminary detection binary signal 242 is submitted to two Add and Remove Conditional Consolidation (ARCC) operators 250 and 259 in series. A complete ARCC operator is given by the following equation: 6 $\begin{array}{cc}\mathrm{out}\left(i,j\right)=\{\begin{array}{cc}1,& \mathrm{if}\sum _{\left(m,n\right)\in W}^{}\mathrm{in}\left(i-m,j-n\right)\ge \mathrm{Threshold1}\\ & \mathrm{and}\forall \left(m,n\right)\in W,\mathrm{YF}\left(i-m,j-n\right)-\mathrm{YF}\left(i,j\right)<\mathrm{Threshold2}\\ & \mathrm{and}\forall \left(m,n\right)\in W,\mathrm{CuF}\left(i-m,j-n\right)-\mathrm{CuF}\left(i,j\right)<\mathrm{Threshold2}\\ & \mathrm{and}\forall \left(m,n\right)\in W,\mathrm{CvF}\left(i-m,j-n\right)-\mathrm{CvF}\left(i,j\right)<\mathrm{Threshold2}\\ 0,& \mathrm{otherwise}.\\ & \end{array}& \left(6\right)\end{array}$

[0052] In this equation, signal YF 202 denotes the filtered version of the noisy luminance input 100 . Similarly, CuF and CvF at 247 u/v correspond to the filtered version of the interpolated chrominance component inputs Cu and Cv 116 u/v as provided by CH-LOSEGANR 115 . The filtering is provided by the 2D low pass filters 243 u/v . Moreover, in the preferred embodiment, for the first ARCC operator 250 , the window dimension is 5×5, threshold1 at 246 is set to 16 and the internal threshold2 is set to 8. For the second ARCC operator 259 , the window dimension is 21×21 as empirically chosen for typical video ITU- 601 signal, the threshold1 at 255 is set to 3 and the internal threshold2 to 8. The second operator output signal 257 represents the Flat (F) region map.

[0053] It is interesting to note that omitting the chrominance components in Equation (6) yields a possible simplified, but less efficient version for Flat region consolidation.

[0054] The Flat region map signal 257 is applied together with the negation of the first ARCC output 256 to an AND gate 262 . The AND gate output 264 represents the corresponding Near-Flat (NF) regions in which mosquito noise is very noticeable for the human vision system (HVS).

[0055] The Texture (T) region in the present embodiment is computed as NOT all of the four (4)-detected regions: E, NE, F and NF. The Texture region map signal 271 can be obtained with a NOT-AND gate 269 with four appropriate corresponding signal inputs: 268 , 251 , 257 and 262 .

[0056] Finally, combining together the five above region maps by the classification block 252 yields the picture Region Map signal 105 utilized for noise power weighting. In order to avoid the potential conflict when a given pixel is classified to more than one region, classification is based on the following priority: Edge, Near Edge, Near Flat, Flat and Texture.

[0057] Referring now to FIG. 3 , there is illustrated a block diagram for the LUminance component REgion-Based Noise power Estimation (LU-REBNE) generally designated at 106 .

[0058] First of all, it can be frequently observed that there is no important signal component in a diagonal high frequency spatial domain. It is thus reasonable to use a diamond shape filter for noise power estimation. Let the noisy decoded luminance signal Y 100 be applied to the diamond shape high pass filter that is composed of a low pass filter 301 whose output 302 is connected to a subtractor 303 . Subtractor 303 subtracts output 302 from luminance Y 100 . The low pass filter 301 is given by the following impulse response: 7 $\begin{array}{cc}\mathrm{d3}\left(i,j\right)=\left[\begin{array}{ccccc}0& 0& 1& 0& 0\\ 0& 2& 8& 2& 0\\ 1& 8& 20& 8& 1\\ 0& 2& 8& 2& 0\\ 0& 0& 1& 0& 0\end{array}\right]/\left(64\right)& \left(7\right)\end{array}$

[0059] The high pass filter output 304 is applied to an absolute value detector 305 whose output is sent in turn to a statistic estimator 307 , which is a shape-adaptive windowing local standard deviation (SD) estimator. The shape adaptive windowing, conceptually based on a homogenous region of similar pixels to the current one in a local window, is required for a reliable local SD estimation in a varying environment in a picture. The shape adaptive windowing segmentation data described further in detail with reference to FIG. 4 , is composed of, at the input 110 , a local binary window, w(i-m, j-n)ε{0,1}, for the current pixel of coordinates (i, j) and, at the input 111 , the number N of “1” for similar pixels to the current pixel in the window. Clearly, N equals to: 8 $\begin{array}{cc}N=\sum _{\left(i,j\right)\in W}^{}w\left(i-m,j-n\right)& \left(8\right)\end{array}$

[0060] A standard deviation estimator, such as 307 , can be generally based on the following equations: 9 $\begin{array}{cc}\mu \left(i,j\right)=\left(\sum _{\left(m,n\right)\in W}^{}w\left(i-m,j-n\right)g\left(i-m,j-n\right)\right)/N\mathrm{and},& \left(9\right)\\ \sigma \left(i,j\right)=C\left(\sum _{\left(m,n\right)\in W}^{}w\left(i-m,j-n\right)g\left(i-m,j-n\right)-\mu \left(i,j\right)\right)/N& \left(10\right)\end{array}$

[0061] In Equation (10), g(i, j), μ(i, j) and σ(i, j) are respectively the estimator input, the internal local mean and the estimated local SD output. Moreover, depending on the anticipated noise distribution the constant C can be chosen in accordance with equal to 1.25 appropriately for additive Gaussian noise, to 1.15 for additive uniform noise or, simply omitted. In the preferred embodiment, the window dimension is chosen as 5 lines×11 columns. For the high frequency signal, the local mean μ(i,j) can be set to zero in Equation (10).

[0062] The SD estimator 307 output, at 308 , is provided to a look-up table SD-LUT 309 further controlled by Coding Parameters 102 . The purpose of SD-LUT 309 is to convert the estimated local standard deviation at 308 to the standard deviation of an equivalent additive noise. SD-LUT 309 generation is previously described in U.S. patent application Ser. No. 09/603,364 (now U.S. Pat. No. 6,633,683 issued Oct. 14, 2003) by the present inventors and assigned to the same assignee, which application is incorporated herein by reference. In that application, a generic configuration and method for random and correlated noise reduction are described. SD-LUT 309 estimates at its output 310 a mean value of local noise input SD σ _m (x, y) (or variance σ _m ² (x,y)). The LUT input-output relationship between the two local standard deviations or σ _r (x, y) (or variance σ _r ² (x, y)) and σ _m (x, y) (or variance σ _m ² (x, y)) can be described by the following method. Let consider the linear portion of the expression representing weight K(x, y):

K ( x, y )=(σ _g ² ( x, y )−σ _n ² ( x, y ))/(σ _g ² ( x, y )) (11)

[0063] wherein the unknown additive noise variance σ _n ² (x, y) is expected to be varying. It is thus necessary to pre-estimate this variance value for each pixel located at (x, y).

[0064] Referring now to FIG. 7 , in many situations where the processing is well defined, such as for NTSC or PAL encoding/decoding and DCT-based compression/decompression, an available original and clean test signal f(x,y) can be used for noise evaluation. FIG. 7 illustrates partly a proposed configuration generally designated at 760 used for performing an off-line noise variance pre-estimation. The original test signal f(x,y) at 750 is applied to the above-mentioned processing at 751 that gives a test noisy image signal g(x,y) at 752 . The additive test noise signal n(x,y) at 754 is then obtained by the difference (g(x,y)−f(x,y)) provided by an adder 753 and is sent in turn to a statistic calculator 755 similar to the calculator 307 shown in FIG. 3 . The test noise SD σ _in (x, y) (or the test noise variance σ _in ² (x,y)) estimation is done in the same context as that of the luminance signal Y 100 in LU-LOSEGANR 108 shown in FIG. 4 , with the segmented window parallel signals w(l-m,j-n) at 110 and the selected-pixels count signal N at 111 . That is, for a considered pixel at (x, y), one may obtain a pair of SD values (σ _r (x, y), σ _n (x, y)) (or a pair of variance values (σ _r ² (x, y),σ _n ² (x, y))). For the whole test picture or set of test pictures, a given value of σ _r (or σ _r ² ) can have many resulting values of σ _in (or σ _in ² ). In order to obtain a unique input-output relationship for the SD-LUT 309 , it is necessary, for a given σ _r (or σ _r ² ), to define a single value σ _in representing all possible values of σ _n . For the preferred SD calculation, proposed estimations for σ _m are as follows:

σ _m =mean (σ _in , given a value of σ _r ) (12)

[0065] or

σ _m =mode (σ _in , given a value of σ _r ) (13)

[0066] The estimation (12) or (13) can be done then on an off-line basis by a data storage and estimation device 757 . The input-output result (σ _r ,σ _m ) at 704 and 758 respectively, permits the establishment of a pre-calculated SD-LUT 309 for real time processing involving an unknown image. If the memory SD-LUT 309 is large enough, some controllable bits can be fed at parameters input 102 representing a learning or functional condition, for example for NTSC, PAL or 12 Mbit MPEG. The main requirement of the method is the prior knowledge of the processing to create the noisy image g(x,y) from the clean image f(x,y). In the present case, the SD-LUT 309 is empirically obtained with various test sequences coded by 16 usual bit rates corresponding to end-user controlled Coding Parameters 102 . FIG. 8 , in the preferred embodiment, represents typically the relationship between the observed SD and the noise coding SD for various Coding Parameters 102 . The SD-LUT output 310 , designated by σ _m (i,j), is applied to the weighting function 311 for MMSE noise reduction explained further with reference to FIG. 4 . Depending on the detected region at the current pixel location (i,j) indicated by region map 105 , the weighting function output signal 312 , designated now by σ _e (i,j), is empirically given by the following equation: 10 $\begin{array}{cc}{\sigma }_e\left(i,j\right)=\{\begin{array}{cc}\left(4/8\right)·{\sigma }_q\left(i,j\right),& \mathrm{for}\mathrm{Edge}\mathrm{Region}\\ \left(5/8\right)·{\sigma }_q\left(i,j\right),& \mathrm{for}\mathrm{Near}\mathrm{Edge}\mathrm{Region}\\ \left(5/8\right)·{\sigma }_q\left(i,j\right),& \mathrm{for}\mathrm{Near}\mathrm{Flat}\mathrm{Region}\\ \left(2/8\right)·{\sigma }_q\left(i,j\right),& \mathrm{for}\mathrm{Flat}\mathrm{Region}\\ \left(2/8\right)·{\sigma }_q\left(i,j\right),& \mathrm{for}\mathrm{Texture}\mathrm{Region}\end{array}& \left(14\right)\end{array}$

[0067] It is worthwhile to note that, in the present embodiment, the noise contribution on Edge pixel is considered as important as the noise contribution on Near-Edge or Near-Flat regions. Such noise will be heavily filtered in these three regions. Inversely, the filtering in Texture region should be sufficiently light enough, since texture already masks noise. Finally, in Flat regions, noise is relatively small and nearly random; excessive filtering will degrade eventually fine but visible signal details.

[0068] In order to smooth the region transitions, the weighting function output signal 312 σ _e (i,j) is applied to a 2D low pass filter L 2 at 313 , which is a separable filter. The 2D impulse response is: 11 $\begin{array}{cc}{L}_2\left(i,j\right)=\left[\begin{array}{ccc}1& 2& 1\\ 2& 4& 2\\ 1& 2& 1\end{array}\right]/\left(16\right)& \left(15\right)\end{array}$

[0069] The filter output signal σ _n (i,j) at 107 , considered as local SD of an equivalent additive but varying noise, is provided to the noise correcting block 108 .

[0070] Referring now to FIG. 4 , there is illustrated a block diagram of the LUminance component LOcal SEGmentation-based Adaptive Noise Reducer (LU-LOSEGANR) 108 . There are many Spatial Adaptive Noise Reduction techniques known in the art. However, few of them are, firstly, robust in presence of noise and, secondly, efficient in the Edge Region(s) of a picture. LU-LOSEGANR 108 is a simplified version of the generic Adaptive Noise Reducer described in the above-cited U.S. patent application Ser. No. 09/603,364. In order to give some robustness to a local segmentation in the presence of noise, a simple low pass filter 401 described by Equation (15) is utilized for the noisy input signal 100 . The filter output 402 , denoted as g*(i,j) is applied to a local window segmentation 403 . The later provides, in the considered window domain W, a set of binary signals w(i-m,j-n) 110 defined as: 12 $\begin{array}{cc}\text{For}\left(m,n\right)\in W,w\left(i-m,j-n\right)=\{\begin{array}{cc}1,& \text{if}g*\left(i-m,j-n\right)-g*\left(i,j\right)<\text{Threshold}\\ 0,& \mathrm{otherwise}\end{array}& \left(16\right)\end{array}$

[0071] Thus, the binary signals w designate a homogenous region, within a threshold tolerance, to the current pixel located at (i,j). The local window therefore becomes shape-adaptive. The threshold value is applied at 416 and is set, in the preferred embodiment, to 12. The number N, at 111 , of “1” for similar pixels in the window, is provided by the counter 405 .

[0072] In order to provide efficient estimation of the two first order signal statistics, the window binary signals 110 and its parameter N signal 111 are connected to a local mean calculator 404 and a local SD calculator 407 for the noisy input signal 100 . The calculators are described respectively again by the above equations (9) and (10).

[0073] Finally, in order to provide efficient noise reduction in a varying environment of picture signal, such as edge regions, a MMSE coring technique is given by a gain calculator 411 operating on the two SD values, the first one σ(i,j) 408 coming from the noisy signal, the second one σ _n (i,j) 107 coming from noise power estimator illustrated in FIG. 3 . The said MMSE coring K(i,j), at 415 , is described by the following equation: 13 $\begin{array}{cc}K\left(i,j\right)=\max \left[0,\frac{{\sigma }^2\left(i,j\right)-{\sigma }_n^2\left(i,j\right)}{{\sigma }^2\left(i,j\right)}\right]& \left(17\right)\end{array}$

[0074] A possible simplified version of Equation (17), at the expense of heavier signal reduction, is an MMSE-like coring defined as: 14 $\begin{array}{cc}K\left(i,j\right)=\max \left[0,\frac{\sigma \left(i,j\right)-{\sigma }_n\left(i,j\right)}{\sigma \left(i,j\right)}\right].& \left(18\right)\end{array}$

[0075] Finally, the filtered output luminance signal Y*(i,j) at 109 is given by

Y *( i,j )=mean[ Y ( i,j )]+ K ( i,j ). {[ Y ( i,j )−mean[ Y ( i,j )]]} (19)

[0076] using a first adder 409 having its output 410 feeding a multiplier 414 receiving K(i,j) at 415 and feeding in turn a second adder 412 , as illustrated in FIG. 4 .

[0077] Referring now to FIG. 5 that represents the CHrominance component LOcal Noise power Estimator (CH-LONE) block diagram 112 of FIG. 1 . The CH-LONE principle is similar to the luminance case. For each chrominance component, as illustrated, CH-LONE 112 comprises a diamond shape low-pass filter 501 for high frequency component extraction with an output 502 connected to a subtractor 503 . Subtractor 503 subtracts output 502 from Noisy Cu/Cv 101 uv for output 504 to an absolute value detector 505 . Local shape adaptive noise power estimator with a 2H up-sampler 507 receives output 506 from detector 505 . 1D low-pass filter 509 receives output 508 and supplies its output 510 to a shape-adaptive local standard deviation estimator 511 . Output 512 of estimator 511 is provided to a 2H down-sampler 513 and thereafter via output 514 to additive noise SD-LUT 515 . SD-LUT output 516 is connected to multiplier 518 that applies a weighting factor 517 . The proposed configuration is based on some assumptions: firstly, for simplicity purpose, the shape adaptive local windowing can