Grid-based quantization-aware compression for high dynamic range images
An apparatus comprising an interface and a processor. The processor may be configured to determine a grid size for a compression operation in response to a source bit-depth and a compressed bit-depth, divide the image frames into grids in response to the grid size, analyze the pixel data for each of the grids to determine a gain and an offset for the grids, process the grids in response to the gain and the offset, calculate a value at the compressed bit-depth for each of the pixel data, store compression parameters in an image header, and generate a compressed image frame comprising the image header and the values for the pixel data at the compressed bit-depth. The compression parameters may comprise a table of the gain and offset for each of the grids. The compression operation may convert the pixel data from the source bit-depth to the compressed bit-depth.
1 . An apparatus comprising:
an interface configured to receive pixel data arranged as image frames; and
a processor configured to (i) determine a grid size for a compression operation in response to a source bit-depth and a compressed bit-depth, (ii) divide said image frames into a plurality of grids in response to said grid size, (iii) analyze a statistical distribution of said pixel data for each of said grids to determine a gain parameter and an offset parameter for each of said grids, (iv) selectively apply processing to each of said grids of said image frames in response to said gain parameter and said offset parameter, (v) calculate a value at said compressed bit-depth for each of said pixel data in said grid, (vi) store compression parameters in an image header, and (vii) generate a compressed image frame comprising said image header and said values for said pixel data at said compressed bit-depth, wherein
(a) said compression parameters comprise said grid size, said source bit-depth and a table comprising said gain parameter and said offset parameter for each of said grids,
(b) said compression operation generates said compressed image frame by converting said pixel data of said image frames from said source bit-depth to said compressed bit-depth, and
(c) said compressed bit-depth is less than said source bit-depth.
2 . The apparatus according to claim 1 , wherein each of said image frames comprise a high-dynamic range image and said compressed image frame comprises a compressed version of said high-dynamic range image.
3 . The apparatus according to claim 1 , wherein said processor is further configured to perform a decompression operation on said compressed image frame from said compressed bit-depth to said source bit-depth.
4 . The apparatus according to claim 3 , wherein said decompression operation comprises (i) parsing said grid size and said compressed bit-depth from said image header, (ii) dividing said compressed image frame into a plurality of decompression grids based on said grid size, (iii) loading said gain parameter and said offset parameter for each of said decompression grids from said image header, and (iv) reconstructing said image frames from said values for said pixel data in each of said decompression grids based on said values for said pixel data, said gain parameter and said offset parameter.
5 . The apparatus according to claim 3 , wherein said processor is configured to achieve a high compression ratio for said compressed image frame while maintaining a fidelity of decompressed data after said decompression operation.
6 . The apparatus according to claim 1 , wherein said grid size comprises an equal width and height.
7 . The apparatus according to claim 1 , wherein said grid size is calculated based on a tradeoff of a compression ratio against (i) an amount of data for representing said compression parameters in said image header and (ii) a system complexity for performing said compression operation.
8 . The apparatus according to claim 1 , wherein said statistical distribution of said pixel data is configured to determine a presentation of said pixel data to enable compact storage and reliable data recovery.
9 . The apparatus according to claim 1 , wherein said compression operation is configured to preserve an integrity of original data of said image frames.
10 . The apparatus according to claim 1 , wherein (i) said compression operation enables a lossy compression and (ii) an amount of loss from said compression operation corresponds to said grid size and a precision of said gain parameter.
11 . The apparatus according to claim 1 , wherein said image frames are represented using said source bit-depth of 20 bits or greater and said compressed image frame is represented using said compressed bit-depth of 12 bits.
12 . The apparatus according to claim 1 , wherein said compression operation comprises a grid-based quantization aware compression.
13 . The apparatus according to claim 1 , wherein said compressed bit-depth is determined in response to analyzing each of said pixel data in one of said grids in real time and adopting a minimum number of bits for pixel representation.
14 . The apparatus according to claim 1 , wherein said compressed bit-depth is determined in response to analyzing a pixel distribution of said image frames offline.
15 . The apparatus according to claim 1 , wherein said gain parameter and said offset parameter for each of said grids reduce redundant data in said image frames while preserving critical details.
16 . The apparatus according to claim 1 , wherein said compression operation are adjusted according to characteristics of said statistical distribution of each of said grids of said image frames to prevent over-compression and over-expansion.
17 . The apparatus according to claim 1 , wherein said value at said compressed bit-depth for each of said pixel data in one of said grids is determined in response to (i) determining a minimum value and a maximum value for each of said pixel data in said one of said grids, (ii) determining said offset parameter based on said compressed bit-depth and said minimum value, (iii) calculating a delta value in response to said maximum value and said offset parameter, (iv) determining said gain parameter in response to said delta value and said compressed bit-depth and (v) calculating said value with said compressed bit-depth in response to said pixel data, said offset parameter and said gain parameter.
18 . The apparatus according to claim 17 , wherein said offset parameter is (i) a maximum integer value for said compressed bit-depth if said minimum value is greater than said maximum integer value for said compressed bit-depth and (ii) said minimum value if said minimum value is not greater than said maximum integer value for said compressed bit-depth.
19 . The apparatus according to claim 17 , wherein (i) said delta value is calculated by subtracting said offset parameter from said maximum value and (ii) said gain parameter is (a) unity if said delta value is less than or equal to a maximum integer value for said compressed bit-depth and (b) said delta value divided by said maximum integer value for said compressed bit-depth if said delta value is greater than said maximum integer value for said compressed bit-depth.
20 . The apparatus according to claim 17 , wherein said value with said compressed bit-depth is calculated by subtracting said offset parameter from a value of said pixel data at said source bit-depth and then dividing by said gain parameter.