Pedestrian feature extraction is one of the key steps in pedestrian recognition. The traditional method of pedestrian recognition is to extract feature descriptors(such as HOG,LBP) from each color channel (R, G, B channels), Finally merge into a feature vector. it is difficult to take account of the correlation information between different color channels. In this paper, we use a holistic approach to extract local feature descriptors from color images, which is called quaternionic local ranking binary pattern local descriptor (QLRBP). Unlike traditional methods, this method extracts LBP features from the quaternionic representation space instead of the three color channels. First, Encoding a color pixel using a quaternion to get the quaternionic representation (QR) of the color image which collected from a vehicle mounted camera. Then, Applying a Clifford translation to QR of the color image. Finally, Performing a local binary codingon the phase of the transformed result to generate local descriptors of the color image. QLRBP is able to handle all color channels directly in the quaternionic domain and include their relations simultaneously. In the method of pedestrian recognition, the positive and negative samples are collected first. The QLRBP features are extracted from all the samples, and the K-nearest neighbor algorithm is used to train the classifier. The method is tested on the INRIA pedestrian database and shows that it is better than other features, such as HOG features and traditional LBP features. Performance approach to the current advanced method of pedestrian recognition.