Image Upsampling using L3 - Revision history

imported>Student2018: /* Appendix */

2018-12-16T17:34:34Z

Appendix

← Older revision		Revision as of 17:34, 16 December 2018
Line 68:		Line 68:
	== Appendix ==		== Appendix ==

	[[Media:Approach_1.zip]]: Matlab script files for the training and rendering steps in Approach 1.		[[Media:Approach_1.zip]]: Matlab script files and training and test scene data for the training and rendering steps in Approach 1.

	== References ==		== References ==

imported>Student2018: /* Appendix */

2018-12-16T17:22:42Z

Appendix

← Older revision		Revision as of 17:22, 16 December 2018
Line 68:		Line 68:
	== Appendix ==		== Appendix ==

	[[Media:Approach_1.zip]]		[[Media:Approach_1.zip]]: Matlab script files for the training and rendering steps in Approach 1.

	== References ==		== References ==

imported>Student2018: /* Appendix */

2018-12-16T17:19:44Z

Appendix

← Older revision		Revision as of 17:19, 16 December 2018
Line 68:		Line 68:
	== Appendix ==		== Appendix ==

	[[~~File~~:Approach_1.zip]]		[[Media:Approach_1.zip]]

	== References ==		== References ==

imported>Student2018: /* Appendix */

2018-12-16T17:19:15Z

Appendix

← Older revision		Revision as of 17:19, 16 December 2018
Line 68:		Line 68:
	== Appendix ==		== Appendix ==

	[[~~Media~~:Approach_1.zip]]		[[File:Approach_1.zip]]

	== References ==		== References ==

imported>Student2018: /* Appendix */

2018-12-16T17:13:40Z

Appendix

← Older revision		Revision as of 17:13, 16 December 2018
Line 68:		Line 68:
	== Appendix ==		== Appendix ==

	[[~~Approach 1~~: ~~Matlab file for the rendering step of the image upsampling algorithm~~]]		[[Media:Approach_1.zip]]

	== References ==		== References ==

imported>Student2018: /* Appendix */

2018-12-16T17:08:47Z

Appendix

← Older revision		Revision as of 17:08, 16 December 2018
Line 68:		Line 68:
	== Appendix ==		== Appendix ==

	[[~~Media~~:~~render.m~~]]		[[Approach 1: Matlab file for the rendering step of the image upsampling algorithm]]

	== References ==		== References ==

imported>Student2018: /* Appendix */

2018-12-16T17:06:27Z

Appendix

← Older revision		Revision as of 17:06, 16 December 2018
Line 67:		Line 67:

	== Appendix ==		== Appendix ==

			[[Media:render.m]]

	== References ==		== References ==

imported>Student2018: /* References */

2018-12-16T17:04:39Z

References

← Older revision		Revision as of 17:04, 16 December 2018
Line 65:		Line 65:

	Once the training is done, the inference or rendering step is quite fast and could be efficiently done in a parallelized way on a GPU. To explore further possibilities with such a scheme, deep neural network architectures could be employed. In such a deep learning scheme, there is no need to predefine various classes as it can learn the features itself much more efficiently.		Once the training is done, the inference or rendering step is quite fast and could be efficiently done in a parallelized way on a GPU. To explore further possibilities with such a scheme, deep neural network architectures could be employed. In such a deep learning scheme, there is no need to predefine various classes as it can learn the features itself much more efficiently.

			== Appendix ==

	== References ==		== References ==

imported>Student2018: /* Approach 2 */

2018-12-14T05:31:25Z

Approach 2

← Older revision		Revision as of 05:31, 14 December 2018
Line 52:		Line 52:
	In this approach, we extend the existing ISET L3 package to include upsampling. Here, the rendering step takes the low resolution sensor data to directly produce the pixel values in the target color space for the higher resolution image. Unlike the previous approach, image processing is included in the rendering step, i.e., the final pixel values are directly generated from the low resolution sensor data. This is done by training kernels that multiply the low resolution sensor data patches to yield RGB values of sets of pixels in the target image instead of the RGB value of just the center pixel. A new class object is introduced, l3DataUpsample() with a property upsample, which is the factor by which the image is upsampled. A new method, l3DataGetUpsampling() is introduced to obtain high resolution sensor data from the input scenes. Further, l3ClassifyFast() is modified to provide the correct target data patches when the upsampling factor is not 1.		In this approach, we extend the existing ISET L3 package to include upsampling. Here, the rendering step takes the low resolution sensor data to directly produce the pixel values in the target color space for the higher resolution image. Unlike the previous approach, image processing is included in the rendering step, i.e., the final pixel values are directly generated from the low resolution sensor data. This is done by training kernels that multiply the low resolution sensor data patches to yield RGB values of sets of pixels in the target image instead of the RGB value of just the center pixel. A new class object is introduced, l3DataUpsample() with a property upsample, which is the factor by which the image is upsampled. A new method, l3DataGetUpsampling() is introduced to obtain high resolution sensor data from the input scenes. Further, l3ClassifyFast() is modified to provide the correct target data patches when the upsampling factor is not 1.

	When upsample =2. similar to the previous approach, the upsampled image is twice the dimensions of the input image. We use a <math>55</math> sized patch, <math>202*4</math> classes corresponding to 20 illuminant levels, 2 contrast levels and 4 color filters in the Bayer pixel array. The algorithm is trained on 5 scenes at 3 exposure times (15 input images). The figure below shows the upsampled image produced by the trained algorithm on a test scene (not present in the training set). The left panel shows the low resolution image, the middle panel is the rendered upsampled image and the right panel is the target higher resolution image.		When upsample = 2, similar to the previous approach, the upsampled image is twice the dimensions of the input image. We use a <math>55</math> sized patch, <math>202*4</math> classes corresponding to 20 illuminant levels, 2 contrast levels and 4 color filters in the Bayer pixel array. The algorithm is trained on 5 scenes at 3 exposure times (15 input images). The figure below shows the upsampled image produced by the trained algorithm on a test scene (not present in the training set). The left panel shows the low resolution image, the middle panel is the rendered upsampled image and the right panel is the target higher resolution image.

	[[File:Fig6_scene6_2.png\|1000px]]		[[File:Fig6_scene6_2.png\|1000px]]

imported>Student2018: /* Results */

2018-12-14T05:30:59Z

Results

← Older revision		Revision as of 05:30, 14 December 2018
Line 50:		Line 50:
	====Approach 2====		====Approach 2====

	In this approach, we extend the existing ISET L3 package to include upsampling. Here, the rendering step takes the low resolution sensor data to directly produce the pixel values in the target color space for the higher resolution image. Unlike the previous approach, image processing is included in the rendering step, i.e., the final pixel values are directly generated from the low resolution sensor data. ~~Similar~~ to the previous approach, the upsampled image is twice the dimensions of the input image. We use a <math>55</math> sized patch, <math>202*4</math> classes corresponding to 20 illuminant levels, 2 contrast levels and 4 color filters in the Bayer pixel array. The algorithm is trained on 5 scenes at 3 exposure times (15 input images). The figure below shows the upsampled image produced by the trained algorithm on a test scene (not present in the training set). The left panel shows the low resolution image, the middle panel is the rendered upsampled image and the right panel is the target higher resolution image.		In this approach, we extend the existing ISET L3 package to include upsampling. Here, the rendering step takes the low resolution sensor data to directly produce the pixel values in the target color space for the higher resolution image. Unlike the previous approach, image processing is included in the rendering step, i.e., the final pixel values are directly generated from the low resolution sensor data. This is done by training kernels that multiply the low resolution sensor data patches to yield RGB values of sets of pixels in the target image instead of the RGB value of just the center pixel. A new class object is introduced, l3DataUpsample() with a property upsample, which is the factor by which the image is upsampled. A new method, l3DataGetUpsampling() is introduced to obtain high resolution sensor data from the input scenes. Further, l3ClassifyFast() is modified to provide the correct target data patches when the upsampling factor is not 1.

			When upsample =2. similar to the previous approach, the upsampled image is twice the dimensions of the input image. We use a <math>55</math> sized patch, <math>202*4</math> classes corresponding to 20 illuminant levels, 2 contrast levels and 4 color filters in the Bayer pixel array. The algorithm is trained on 5 scenes at 3 exposure times (15 input images). The figure below shows the upsampled image produced by the trained algorithm on a test scene (not present in the training set). The left panel shows the low resolution image, the middle panel is the rendered upsampled image and the right panel is the target higher resolution image.

	[[File:Fig6_scene6_2.png\|1000px]]		[[File:Fig6_scene6_2.png\|1000px]]