This paper explaining different approaches starting from 1960’s. Face recognition (FR) has been broadly studied by several authors over the last thirty years. Different feature extraction methods with its previous work are discussed. Many approaches have been taken which has lead to different algorithms. Some of the most relevant are PCA, ICA, LDA and their derivatives We may classify these approaches as better, less accurate, computationally costly, time consuming, acceptable to some extent. Although face detection technology is now sufficiently mature to meet the minimum requirements of many practical applications, much work is still needed before automatic face detection can achieve performance comparable to the human performance. However, still automatic face recognition system faces some problems due to environmental conditions and constraints of hardware, light sources going to use.Many algorithms use the color information as a feature (some of them may be gray-scale) to recognize faces. The color that we perceive not only depends on the surface’s nature, but also the light on it. The intensity level of a pixel and variation/ relation between the pixels can vary greatly depends on the lighting conditions. Fig. 8 shows the illumination conditions on same image taken under uncontrolled environment. As many feature extraction methods used intensity variance measure between pixels to obtain relevant data, researchers show an important dependency on lighting changes. If new light sources are added then the light intensities may increase or decrease. Because of solar light, entire face regions be obscured or in shadow. . The big problem is that two faces of the same subject with illumination variations may show more differences between them. By considering all above points, illumination is one of the big challenges of automated face recognition systems. However, it has been demonstrated that humans can generalize representations of a face under radically different illumination conditions, although human recognition of faces is sensitive to illumination direction 148, 149.The Haar + AdaBoost approach is effective and efficient. However, the current approach has almost reached its power limit. Within such a framework, possible improvements may be possible by designing additional sets of features that are complementary to the existing ones and adopting more advanced learning techniques, which could lead to more complex classifiers while avoiding the over fitting problem.However, in almost all of the cases, ambient illumination is the foremost challenge for most face recognition applications. The accuracy of face recognition systems highly depends on the features that are extracted to represent the face which, in turn, depends on correct face localization and normalization. A powerful classification engine is still necessary to deal with difficult nonlinear classification and regression problems in the constructed feature space. Feature-based techniques were not effective.Geometric properties alone are inadequate for face recognition because rich information contained in the facial texture or appearance is not utilized. These are the main reasons why earl Face recognition technology has made impressive gains, but it is still not able to meet the accuracy of many applications. A sustained and collaborative effort is needed to address many of the open problems in face recognition. Some papers deals with illumination invariance, showing interesting results 144, 145. LDA shows better performance than PCA. Bayesian methods allow to define intra-class variations such as illumination variations. However, all this linear analysis algorithms do not capture satisfactorily illumination variations. Non-linear methods handle the illumination changes better than linear methods. Kernel PCA and non-linear SVD shows better performance than linear methods. Choosing the right statistical model can help to deal with illumination issues efficiently. Nevertheless, despite the progress made in last years and the different solutions proposed, FR performance is not yet satisfactory when more demanding conditions are required (different viewpoints, blocked effects, , strong lighting states, etc). As a consequence great progress has been achieved toward developing computer vision algorithms that can recognize individuals based on their facial images in a similar way that human beings do, and leading this technology to reliable personal identification systems. This has been possible due to the increase of computational power of state-of-the-art computers. Nevertheless, in real and non-controlled environments, FR systems still remain an open challenge and major problems remain to be solved.Aside from the above statements, advantages of FR automated systems cannot be dismissed. This technology is turning more popular when compared with other biometric modalities. Thus, unlike iris, retinal, hands-geometry or fingerprint recognition systems, FR does not require high accuracy and expensive image acquisition equipments.