Final Report Summary - POCS-DSR (POCS based Depth Super Resolution)
Publishable Summary
POCS Based Depth Super-Resolution (POCS-DSR) project aims to increase the depth resolution within the working volume of a range sensing camera. Developed algorithms are tested on Time-of-Flight cameras in the first phase, and extended to Light Detection and Ranging (LIDAR) sensors in the second period of the project. The research and development efforts during the project resulted in the following outputs
1. Real-time image registration on DM3730 embedded DSP platform using an innovative feature based alignment technique. [4]
2. Multi-exposure depth enhancement algorithm is devised to capture and fuse valuable depth information in the scene using alternating exposure durations [1],[6],[7].
3. Efficient super-resolution image reconstruction using Projection Onto Convex Sets
a. Parallel image reconstruction algorithms are developed using OpenMP.
b. Graphic processor unit (GPU) implementation using OpenCL & CUDA [2],[3]
c. Porting the OpenCL code to Altera FPGA platform.
4. Application of the developed resolution enhancement technology to biomedical field to enhance diffusion tensor imaging (DTI) [5].
5. Moving object detection algorithm for aerial platforms with false alarm reduction using optical flow. Patent filed [8].
6. Auto-focus algorithm for cameras using longest increasing subsequence (LIS) technique. Patent filed [9].
7. Video contrast enhancement algorithm for low power processors by sparse sampling the original histogram with the help of a massively parallel coprocessor. Patent filed [10].
8. Submitted a journal paper that discusses the problem of constrained interest point selection [11].
Researcher achieved following goals during the project:
• Published several international and local conference papers as a result of the successful integration of the researcher to the host organization.
• Contributed to the writing of Suicide Bomber Counter Action and Prevention (SUBCOP) proposal with a consortium of researchers from 11 companies across Europe. Researcher is currently the principal investigator of the SUBCOP FP7 project within ASELSAN.
• Developed real-time pedestrian detection algorithms.
List of Publications and Inventions During Project:
1. “A System and Method for Resolution Enhancement”, June 20 2013, Publication No WO/2013/088197, Lütfi Murat Gevrekci (TR) ASELSAN
2. Toygar Akgün, Murat Gevrekci, “Accelerating Super-Resolution Using GPU by CUDA”, Computational Science, Engineering and Information Technology (CCSEIT - 2013), June 2013.
3. Toygar Akgün, Murat Gevrekci, “GPU Based Resolution and Contrast Enhancement for Infrared Cameras”, NVIDIA GPU Technology Conference (GTC), March 2013
4. Murat Gevrekci, Umut Demircin, Erdem Akagündüz, “Real-time image registration”, IEEE Signal Processing and Communications Applications Conference, pp. 1-4, 2012.
5. Özgür Yılmaz, Murat Gevrekci, Hüseyin Boyacı, Katja Doerschner, “Super-resolution in diffusion tensor imaging”, Turkish Magnetic Resonance Association 16. Annual Meeting, May 2011.
6. Murat Gevrekci, Kubilay Pakin, “Depth Map Super-resolution”, IEEE Int. Conf. on Image Processing (ICIP), pp.3449-3452 September 2011.
7. Murat Gevrekci, Kubilay Pakin, “Depth Map Super-resolution”, Signal Processing and Communications Applications Conference, pp.502-505 2011.
8. Murat Gevrekci, Mehmet Umut Demircin, Erkan Okuyan, “Moving Object Detection Apparatus and Method using Optical Flow History Imaging False Alarm Reduction,” (patent filed March 24, 2015)
9. Erkan Okuyan, Murat Gevrekci, Berk Ülker, Cevdet Aykanat “System and Method for Optimizing Focus of an Optical System Passively”. (patent filed Agust 18, 2015)
10. Murat Gevrekci, Hüseyin Seçkin Demir, Buse Gül Atlı, Hakan Atbaş, Cihan Alkan, Buket Aykanat, Kutalmış Gökalp İnce. “Sparse Sampling Video Contrast Enhancement Apparatus and Method”, (patent filed July 20, 2015)
11. Erkan Okuyan, Murat Gevrekci, Cevdet Aykanat “Fast constrained interest point selection”, Submitted to Journal of RealTime Image Processing
POCS Based Depth Super-Resolution (POCS-DSR) project aims to increase the depth resolution within the working volume of a range sensing camera. Developed algorithms are tested on Time-of-Flight cameras in the first phase, and extended to Light Detection and Ranging (LIDAR) sensors in the second period of the project. The research and development efforts during the project resulted in the following outputs
1. Real-time image registration on DM3730 embedded DSP platform using an innovative feature based alignment technique. [4]
2. Multi-exposure depth enhancement algorithm is devised to capture and fuse valuable depth information in the scene using alternating exposure durations [1],[6],[7].
3. Efficient super-resolution image reconstruction using Projection Onto Convex Sets
a. Parallel image reconstruction algorithms are developed using OpenMP.
b. Graphic processor unit (GPU) implementation using OpenCL & CUDA [2],[3]
c. Porting the OpenCL code to Altera FPGA platform.
4. Application of the developed resolution enhancement technology to biomedical field to enhance diffusion tensor imaging (DTI) [5].
5. Moving object detection algorithm for aerial platforms with false alarm reduction using optical flow. Patent filed [8].
6. Auto-focus algorithm for cameras using longest increasing subsequence (LIS) technique. Patent filed [9].
7. Video contrast enhancement algorithm for low power processors by sparse sampling the original histogram with the help of a massively parallel coprocessor. Patent filed [10].
8. Submitted a journal paper that discusses the problem of constrained interest point selection [11].
Researcher achieved following goals during the project:
• Published several international and local conference papers as a result of the successful integration of the researcher to the host organization.
• Contributed to the writing of Suicide Bomber Counter Action and Prevention (SUBCOP) proposal with a consortium of researchers from 11 companies across Europe. Researcher is currently the principal investigator of the SUBCOP FP7 project within ASELSAN.
• Developed real-time pedestrian detection algorithms.
List of Publications and Inventions During Project:
1. “A System and Method for Resolution Enhancement”, June 20 2013, Publication No WO/2013/088197, Lütfi Murat Gevrekci (TR) ASELSAN
2. Toygar Akgün, Murat Gevrekci, “Accelerating Super-Resolution Using GPU by CUDA”, Computational Science, Engineering and Information Technology (CCSEIT - 2013), June 2013.
3. Toygar Akgün, Murat Gevrekci, “GPU Based Resolution and Contrast Enhancement for Infrared Cameras”, NVIDIA GPU Technology Conference (GTC), March 2013
4. Murat Gevrekci, Umut Demircin, Erdem Akagündüz, “Real-time image registration”, IEEE Signal Processing and Communications Applications Conference, pp. 1-4, 2012.
5. Özgür Yılmaz, Murat Gevrekci, Hüseyin Boyacı, Katja Doerschner, “Super-resolution in diffusion tensor imaging”, Turkish Magnetic Resonance Association 16. Annual Meeting, May 2011.
6. Murat Gevrekci, Kubilay Pakin, “Depth Map Super-resolution”, IEEE Int. Conf. on Image Processing (ICIP), pp.3449-3452 September 2011.
7. Murat Gevrekci, Kubilay Pakin, “Depth Map Super-resolution”, Signal Processing and Communications Applications Conference, pp.502-505 2011.
8. Murat Gevrekci, Mehmet Umut Demircin, Erkan Okuyan, “Moving Object Detection Apparatus and Method using Optical Flow History Imaging False Alarm Reduction,” (patent filed March 24, 2015)
9. Erkan Okuyan, Murat Gevrekci, Berk Ülker, Cevdet Aykanat “System and Method for Optimizing Focus of an Optical System Passively”. (patent filed Agust 18, 2015)
10. Murat Gevrekci, Hüseyin Seçkin Demir, Buse Gül Atlı, Hakan Atbaş, Cihan Alkan, Buket Aykanat, Kutalmış Gökalp İnce. “Sparse Sampling Video Contrast Enhancement Apparatus and Method”, (patent filed July 20, 2015)
11. Erkan Okuyan, Murat Gevrekci, Cevdet Aykanat “Fast constrained interest point selection”, Submitted to Journal of RealTime Image Processing