Energy-Efficient Change Detection Approach Based on Approximate Computing

The process of image change detection involves the identification and examination of variations or disparities within an image when compared to one or more other images. This method finds extensive application across diverse domains, such as video surveillance, environmental monitoring, and the analysis of satellite imagery. Within the framework of this research paper, we introduce an image change detection technique centered on Discrete Cosine Transform (DCT) with an emphasis on frequency analysis. Our approach introduces two distinct variants of approximate DCT, namely, the coefficient-optimized ADCT-C and the further pixel-optimized ADCT-P. The experimental findings unveil compelling advantages, with the approximate ADCT-C demonstrating a notable reduction of 60.86% in area requirements and a 64.77% decrease in power consumption when contrasted with the precise DCT. Furthermore, we employ the proposed approximate DCT variants in tandem with the change detection algorithm presented in this paper. Comparative evaluations with alternative approximate DCT methods highlight the efficacy of our approach. Notably, ADCT-C outperforms its counterparts across all three image tests, exhibiting the highest Peak Signal-to-Noise Ratio (PSNR) values.