High Performance O(1) Histogram Construction Technique Based on Parallel Data Structure

• Main Authors: Fan-Chieh Cheng, 鄭帆捷
• Other Authors: Shanq-Jang Ruan
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/78546369550527171110

博士 === 國立臺灣科技大學 === 電子工程系 === 100 === Image processing techniques are currently developing for many multimedia devices, such as the monitor, Tablet PC, smart phone, Personal Digital Assistant (PDA), and digital camera. In many image processing applications, the statistical data of neighbors can be utilized for each pixel. However, the exhaustive search of neighborhood extremely increases the additional cost of the computation. Therefore, this thesis proposes a very efficient O(1) histogram construction method by using the spatial redundancy. For each row, the constructed histogram is calculated across adjacent columns. After the scanned line moves downward, the histogram can be updated in constant time using the spatial difference between rows. Furthermore, all elements of the proposed histogram can be directly regarded as the kernel histogram of the image processing. Compared with other state-of-the-art methods, the histogram constructed by the proposed method can be easily employed in parallel due to the data independence. Hence, this thesis also describes the parallel implementation of the proposed method. Experimental results demonstrate that the efficiency of the proposed method in contrast with those of other methods for histogram construction and image processing applications.