An improved face recognition algorithm using adjacent pixel intensity difference quantization

In our previous research, we have proposed a very simple yet highly reliable face recognition algorithm using Adjacent Pixel Intensity Difference Quantization (APIDQ) histogram. We focus on the quantization method of APIDQ. We found that the quantization of the intensity variation vectors directly in (dIx, dIy) rectangular coordinate plane outperforms that in r-9 polar coordinate plane. In this paper, we present an improved quantization method to realize better recognition performance. After the intensity variation vectors for all the pixels in a facial image are calculated, each vector is quantized directly in (dIx, dIy) plane instead of r-Q plane. By counting the number of elements in each quantized area in the (dIx, dIy) plane, a histogram can be created. This histogram, obtained by APIDQ for facial images, is utilized as a very effective personal feature. Experimental results show that proposed face recognition algorithm achieves more reliable recognition performance by using publicly available face database of AT&T Laboratories Cambridge and very large face database of FERET respectively.