No abstract

In this paper, we propose a new texture descriptor GLBP (Gradient LBP) for age group estimation. LBP exploits the only signs of the gradients between a center pixel and its surrounding pixels in a local patch and it does not reflect their magnitudes. This fails LBP to describe local signal structure in detail. Motivated by this observation, we propose so called GLBP, which considers the magnitude as well as the sign of the gradient. Experimental results show that when the proposed method is applied to age group estimation, it can achieve higher classification rate than existing well-known texture descriptors.

SUMMARYSupport Vector Machine (SVM) is one of the most widely used classifiers to categorize observations. This classifier deterministically selects a class that has the largest score for a classification output. In this letter, we propose a multiclass probabilistic classification method that reflects the degree of confidence. We apply the proposed method to age group classification and verify the performance.

The fairly recent standard of equipping mobile devices with advanced imaging sensors has opened the possibility of conveniently diagnosing skin conditions, anywhere, anytime. For this application, we attempted to estimate skin conditions from a skin image taken by a mobile handheld camera. To estimate the skin conditions, we specifically identified three skin features (pigmentation, pores, and roughness) that can be measured quantitatively from a skin image. The experimental data indicate that the existing thresholding methods are inappropriate for extracting the pigmentation and pore skin features. Thus, we propose a new line‐fitting based thresholding method for skin feature detection. We thoroughly evaluated our proposed skin condition estimation method using our skin image database. The experimental results show that our proposed thresholding method can better determine the threshold leading to the most visually plausible detection, when compared to existing methods. We also confirmed that skin conditions can be feasibly estimated using a common mobile handheld camera (for example, a smartphone).