Abstarct: Plants play an important role in human life that has many different kinds where recognition of them can be very useful and called Leaf Recognition System. To now, this process is examined with expert botanists manually, where this acts is a time-consuming, exhausting, errorable and human memory-baxsed case with low-efficiency result. Hence, researches try to implement this process with artificial intelligence algorithms and improve accuracy. According to study, each plant can be identificate with leaves properties, so we collect and divide helpful leaves features into three categories: First category called global features and contains leaf-dimensional, convex area, etc. second category or local features contains texture information and vein structures. Finally third category consists of counter-baxsed features about leaf image. These feature sets experiment in many approaches and report different results in attached papers, but efficiency of these works depended on special conditions and limited leaf-types. Therefore, proposed method defines novel features in each category and then combines above three feature sets. It leads to attain a general leaf recognition system. In order to feature extraction step, proposed local features contain 6 features baxsed on texture mathematical theory that certainly should apply on segmented part of leaf without any teeth and primary vein. Proposed global features take away from recent researches and aggregate in 4 features. Finally, counter-baxsed category consists of novel method for feature extraction using Wavelet transform that produce 4 features. Altogether, features vector with 14 elements has been classify with K-Nearest Neighbour algorithm. Results show that proposed method can successfully answer for total leaf-types and various environment conditions.