Dynamic range compression of output channel data of an image sensor
View Patent ↗A method (S 100 ) for dynamic range compression of output channel data from an image sensor ( 2 ) comprising an array of sensor cells. The method (S 100 ) comprises selecting a window (S 130 ) in the channel data, the window having a reference pixel value and a plurality of nearby pixel values. The reference pixel value originates from a reference cell that is one of the sensor cells and the nearby pixel values originate from the sensor cells that are in close proximity to the reference cell. There is a step of multiplying (S 140 ) the pixel values, in the window, by a respective weight value to provide weighted pixel values and then adding (S 150 ) the weighted pixel values to provide a convolution value. Thereafter, there is a step of providing (S 160 ) a dynamic range compression value for the window from a selected one of the pixel values and said convolution value and then an assigning step (S 170 ) assigns the dynamic range compression value to a selected pixel location comprising part of an image.
1. According to one aspect of the invention there is provided a method for dynamic range compression of output channel data from an image sensor comprising an array of sensor cells, the method comprising the steps of:
selecting a window in the channel data, the window having a reference pixel value and a plurality of nearby pixel values, said reference pixel value originating from a reference cell that is one of the sensor cells and said nearby pixel values originating from said sensor cells that are in close proximity to said reference cell;
multiplying said pixel values in the window by a respective weight value to provide weighted pixel values;
adding said weighted pixel values to provide a convolution value;
providing a dynamic range compression value for said window from a selected one of said pixel values and said convolution value; and
assigning the dynamic range compression value to a selected pixel location comprising part of an image.
2. A method for dynamic range compression as claimed in claim 1 , wherein the step of providing is effected by dividing the said selected one of said pixel values by the convolution value to provide said dynamic range compression value.
3. A method for dynamic range compression as claimed in claim 2 , wherein said selected one of said pixel values is said reference pixel value.
4. A method for dynamic range compression as claimed in claim 1 , wherein the widow corresponds only to a chrominance channel.
5. A method for dynamic range compression as claimed in claim 1 , wherein the steps are repeated so that the step of assigning thereby assigns a said dynamic range compression value to every pixel location providing a luminance color spectrum for said image.
6. A method for dynamic range compression as claimed in claim 1 , wherein said reference cell may is a centrally located sensor cell in an array of said sensor cells providing said nearby pixel values.
7. A method for dynamic range compression as claimed in claim 1 , wherein said step of multiplying is further characterised by the respective weight value having the effect of Gaussian filtering.
8. A method for dynamic range compression as claimed in claim 7 , wherein the step of multiplying is further characterised by all said pixel vales in the window being multiplied by a said respective weight value.
9. A method for dynamic range compression as claimed in claim 1 , wherein said dividing has the effect of non-linear filtering said pixels values in said window.
10. A method for dynamic range compression as claimed in claim 1 , wherein the steps of adding, providing and assigning are effected by:
i
ref
A
+
W
⊗
I
where i ref is the reference pixel value; I is a matrix representing the pixel values in the window for the luminance channel; W is a weight matrix representing strengths of connections for a corresponding sensor cell in the matrix I; and A is matrix containing decay values of the corresponding cell in matrix I; and is a convolution function.
11. A digital camera comprising: an image sensor comprising an array of sensor cells; and a Channel processing circuit coupled to receive channel output data from the image sensor, wherein in use the Channel processing circuit selects a window in the channel data, the window having a reference pixel value and a plurality of nearby pixel values, said reference pixel value originating from a reference cell that is one of the sensor cells and said nearby pixel values originating from said sensor cells that are in close proximity to said reference cell; and thereafter assigns a dynamic range compression value to a selected pixel location comprising part of an image, wherein the Channel processing circuit determines the dynamic range compression value by: multiplying said pixel values in the window by a respective weight value to provide weighted pixel values; adding said weighted pixel values to provide a convolution value; and dividing a selected one of said pixel values by said convolution value.
12. A digital camera as claimed in claim 11 , wherein said selected one of said pixel values is said reference pixel value.
13. A digital camera as claimed in claim 11 , wherein the widow corresponds only to a chrominance channel.
14. A digital camera as claimed in claim 11 , wherein said multiplying is further characterised by the respective weight value having the effect of Gaussian filtering.
15. A digital camera as claimed in claim 11 , wherein said Channel processing circuit effects the function:
i
ref
A
+
W
⊗
I
where i ref is the reference pixel value; I is a matrix representing the pixel values in the window for the luminance channel; W is a weight matrix representing strengths of connections for a corresponding sensor cell in the matrix I; and A is matrix containing decay values of the corresponding cell in matrix I; and is a convolution function.