There have been many researches being done about pedestrian detection, as pedestrians are exposed to lot of danger on the street according to the fatality statistics. To increase the well-being of the pedestrians there has been many breakthroughs in computer vision researches and its application in actual vehicles. However, the reliability and the complexness of algorithm are still very much in doubt to improve the safety of pedestrian. Most computer vision-based algorithms are complex in nature which will in-turn demand the cost issue for the vehicle manufacturers, while most algorithms' reliability drops significantly in night time (dark surroundings), raining or both conditions. In this work, we try to capture the reliability of our "eyes" and the accuracy of computer vision by implementing neuromorphic visual information processing. Our neuromorphic visual information processing uses basis of result from Hubel and Wiesel's experiment on cat's striate cortex. From the result, we have formed our unique neuromorphic visual information processing which can mimic the response given by the cat's striate cortex to the same input. And since its neuromorphic implementation, it is possible to implement it with relatively simple CMOS technology which would reduce the cost in possible future implementation. The algorithm operates by extracting different directions of orientation features using our neuromorphic visual information processing then neural network is applied to it with a template that resembles upper torso of human body. Our method of orientation feature extraction and the usage of neural network ensure the reliability of our algorithm to perform well under other conditions such as in grey surroundings (dusk environment). Current results from data captured in various places in Korea, in London and using the Daimler database shows that our algorithm performs successfully. Those sets of data all varies as the surrounding is different. Especially, the dataset captured in London is with the narrow roads and heavier presence of pedestrians or cyclists. However, despite all this, the detection was successful, as only few changes to parameters could meet the change in the image type. In conclusion, this research has successfully detected pedestrian by using the CMOS neuromorphic approach so as to reduce the cost for implementation whilst improving on the accuracy and reliability of the detection, over 90 % for the video dataset captured during both day and night. Currently there is also an on-going project with Hyundai Motor Co., to implement a hardware that uses our work to detect pedestrians so as to warn them when Hyundai's electric vehicle is near them.