High Dynamic Range Image Rendering With a Retinex-Based Adaptive Filter
L.Meylan,S.Susstrunk, IEEE transactions on Image Processing, vol.15,no.9,Sep 2009
This Paper focuses on reproducing the HDR(High Dynamic Range)images on LDR(Low Dynamic Range) display devices. Processing the HDR image globally may lead to loss of contrast, which leads to loss of detail visibility. Local processing helps to improve the contrast of the image. Local processing along with global compression is necessary to process a HDR image. Global compression scales the image's dynamic range to the output image's dynamic range.
Surround based retinex algorithm is preferred , where the surround replaces the traditional circular shape by an adaptive surround.
Surround based retinex algorithm is preferred , where the surround replaces the traditional circular shape by an adaptive surround.
Explanation of Proposed work:
(1) The method uses parallel processing, where the chrominance and luminance channel are processed in parallel but the retinex-based adaptive filter method is applied to luminance channel only.
(2)The luminance channel is calculated from the first principal component analysis of the input image.
(3)Global compression is applied to the luminance channel and to the linear RGB image.
(4)Retinex based adaptive filter is applied in the log domain of the globally corrected luminance. While Logarithm is applied to the globally corrected RGB image.
(5) PCA transform is applied to transform into chrominance luminance channel.
(6) The first component is replaced by the treated luminance and the image is transformed back to RGB .
Steps :
Global Tone Mapping: It helps in compression of dynamic range
Local Adaptation: It is done with the help of surround based retinex algorithm, where the surround replaces the traditional circular shape by adaptive filter. Adaptive filter adapts to the shape and size thus preventing halo artifacts.
Edge Detection: Canny edge detector is used to detect the high contrast edges.
Post Processing- Histogram Scaling: The outliers are removed by histogram scaling and clipping.
This method increases the local contrast of the image while preventing the halo artifacts.
(1) The method uses parallel processing, where the chrominance and luminance channel are processed in parallel but the retinex-based adaptive filter method is applied to luminance channel only.
(2)The luminance channel is calculated from the first principal component analysis of the input image.
(3)Global compression is applied to the luminance channel and to the linear RGB image.
(4)Retinex based adaptive filter is applied in the log domain of the globally corrected luminance. While Logarithm is applied to the globally corrected RGB image.
(5) PCA transform is applied to transform into chrominance luminance channel.
(6) The first component is replaced by the treated luminance and the image is transformed back to RGB .
Steps :
Global Tone Mapping: It helps in compression of dynamic range
Local Adaptation: It is done with the help of surround based retinex algorithm, where the surround replaces the traditional circular shape by adaptive filter. Adaptive filter adapts to the shape and size thus preventing halo artifacts.
Edge Detection: Canny edge detector is used to detect the high contrast edges.
Post Processing- Histogram Scaling: The outliers are removed by histogram scaling and clipping.
This method increases the local contrast of the image while preventing the halo artifacts.