Multi-scale image normalization and enhancement
View Patent ↗An image may be processed to normalize and/or remove noise from the image. The processing of the image may involve decomposition of the image into multiple components and subsequent gray scale registration across multiple scales.
1. A method of processing an image, comprising:
normalizing the image, including resizing the image and adjusting the image bit depth, to obtain a pre-normalized image;
obtaining from the pre-normalized image, by an automated processing device, multiple component images, using multi-scale decomposition;
performing gray-scale registration based on the multiple component images, wherein the gray-scale registration includes performing histogram matching of the multiple component images in accordance with a target distribution corresponding to at least one histogram obtained based on at least one other image;
removing unwanted information from the image using the multiple component images;
normalizing the image based on the multiple component images; and
enhancing the image by combining a result of the normalizing with a coarse residual component comprising low-frequency information, wherein the combining includes:
multiplying the result of the normalizing by a first factor computed based on a quantized and scaled version of the coarse residual component and on a quantized and scaled version of the result of the normalizing; and
multiplying the coarse residual component by a scalar design parameter for dynamic range reduction.
2. The method of claim 1 , wherein the multi-scale decomposition comprises at least one decomposition technique selected from the group consisting of: steerable pyramids, Laplacian pyramids, wavelets, and curvelets.
3. The method of claim 1 , wherein the multi-scale decomposition results in an upper-frequency sub-image, and wherein removing unwanted information comprises removing at least one component corresponding to the upper-frequency sub-image.
4. The method of claim 1 , wherein performing a multi-scale decomposition results in a set of multi-scale components that includes the coarse residual component.
5. The method of claim 4 , further comprising performing, on the set of multi-scale components, except for the coarse residual component, at least one of said gray scale registration or said removing unwanted information.
6. The method of claim 1 , further comprising downloading software instructions that, if executed by a processor, cause the processor to perform said normalizing to obtain a pre-normalized image, said obtaining, said performing, said removing, said normalizing based on the multiple component images, and said enhancing.
7. A non-transitory computer-readable medium containing software instructions that, if executed by a processor, cause the processor to perform operations comprising:
normalizing the image, including resizing the image and adjusting the image bit depth, to obtain a pre-normalized image;
obtaining from the pre-normalized image, by an automated processing device, multiple component images, using multi-scale decomposition;
performing gray-scale registration based on the multiple component images, wherein the gray-scale registration includes performing histogram matching of the multiple component images in accordance with a target distribution corresponding to at least one histogram obtained based on at least one other image;
removing unwanted information from the image using the multiple component images;
normalizing the image based on the multiple component images; and
enhancing the image by combining a result of the normalizing with a coarse residual component comprising low-frequency information, wherein the combining includes:
multiplying the result of the normalizing by a factor computed based at least in part on a quantized and scaled version of the coarse residual component and on a quantized and scaled version of the result of the normalizing; and
multiplying the coarse residual component by a scalar design parameter for dynamic range reduction.
8. The medium of claim 7 , wherein the multi-scale decomposition comprises at least one decomposition technique selected from the group consisting of: steerable pyramids, Laplacian pyramids, wavelets, and curvelets.
9. The medium of claim 7 , wherein the multi-scale decomposition results in an upper-frequency sub-image, and wherein removing unwanted information comprises removing at least one component corresponding to the upper-frequency sub-image.
10. The medium of claim 7 , wherein performing a multi-scale decomposition results in a set of multi-scale components that includes the coarse residual component.
11. The medium of claim 10 , further comprising performing, on the set of multi-scale components, except for the coarse residual component, at least one of said gray scale registration or said removing unwanted information.
12. The medium of claim 7 , wherein said factor is further computed based on a set of scalar parameters configured to control relative contributions of contrast enhancement.
13. The method of claim 1 , wherein said factor is further computed based on a set of scalar parameters configured to control relative contributions of contrast enhancement.