Pedestrian detection is challenging in infrared images because of the low signal-to-noise ratio (SNR), low contrast, lack of shape, and complex background. In this paper, an adaptive pedestrian detection method based on human visual mechanism and support vector machine is proposed to overcome these difficulties. As per the human visual attention mechanism, firstly mean and Laplacian of Gaussian (LoG) filter is employed to suppress background noise and increases the SNR value, but it increases contrast relation between the foreground and background. In addition, to remove the remaining noise, the morphological process is applied on the filtered image. The pedestrians and non-pedestrians are obtained by applying local thresholding segmentation to the morphological processed image. Further, to recognize the true pedestrian and to reduce the false alarm rate, support vector machine classifier is used. Experiments are carried out on the standard OTCBVS-BENCH-thermal collection over the OSU thermal pedestrian database by comparing with other methods. The proposed method shows the efficacy by high pedestrian detection rate and low false alarm rate. The contributions of this research work are as follows: (1) The parameters required for the processing of proposed method is calculated adaptively, (2) Background suppression and enhancement of pedestrian is done using human visual system property contrast mechanism, i.e., LoG filter with kurtosis, (3) The pedestrians and non-pedestrians are detected using L-moment-based local thresholding, and (4) The proposed method can work for an input image of any size with similar time complexity. © 2017, King Fahd University of Petroleum & Minerals.