Energy-Efficient Approximate Compressor Design for Image Processing Applications

In microelectronics, the primary objective is to design a circuit with minimal circuit area and less hardware complexity. Furthermore, there should be minimal power consumption and propagation delays. However, in applications such as image processing, the accuracy of the output is not of significant importance and can deviate significantly. The core objective of this work is to join a thread called approximate computing. A novel 4:2 approximate compressor is indended with a novel approximation based full adder. The proposed design is compared with the exact and three approximate designs, and its performance is validated. Moreover, the circuits are simulated with 16 nm technology. In addition, a practical evaluation of approximate designs is performed and edge detection is computed to validate the proposed method. Performance metrics such as the Peak Signal-to-Noise Ratio (PSNR), Mean Squared Error (MSE), Mean Absolute Error (MAE) and Structural Similarity Index Measure (SSIM) are computed for the analysis of image processing applications with the help of Python code in the Google Colab platform.