Accurate detection of bright and dark lesions plays a very important role in diagnosis of various retinal related pathologies. This paper presents a method for detection of exudates and microaneurysms (MAs) from the retinal images by leveraging the property of high spatial frequencies of local pixel intensity. In contribution, we proposed a method that use delta local rank transform with positive and negative delta values to detect both exudates and MAs respectively. The exudate candidates are obtained after retinal anatomic structure removal by masking optic disc (OD) and blood vessels. Finally, selecting the bright and dark candidates by filtering out noise and postprocessing techniques. To demonstrate our proposed method, we carried out an experiments to detect lesion on two publicly available datasets namely, E-Ophtha and Diaretdbl. The result proves that the proposed method is comparable with the other state-of-the-art methods. © 2018 IEEE.