25-08-2014, 02:48 PM
MODALITY REDUCTION AND FACE SKETCH RECOGNITION PROJECT REPORT
MODALITY REDUCTION.docx (Size: 102.22 KB / Downloads: 10)
ABSTRACT
Automatic face sketch recognition plays an important role in law enforcement. Recently, various methods have been proposed to address the problem of face sketch recognition by matching face photos and sketches, which are of different modalities. However, their performance is strongly affected by the modality difference between sketches and photos. In this project, we will first try to reduce the modality difference between face photos and face sketches by transforming images into feature domain. In our system, face sketch image is the input image and output is the face photo corresponding to the input face sketch image. The system that we would like to develop has five components: 1) extract facial components from face photos and face sketch images 2) compute the local binary pattern for each facial component 3) extract curvelet features from face photo and face sketch 4) apply some classifier (like minimum distance classifier or ANN) in order to get best matching face photos corresponding to the probe face sketch 5) display the face photo which is best matching with input face sketch.
INTRODUCTION
Face recognition from still images and video sequences is emerging as an active research area over last 30 years with numerous commercial and law enforcement application. Face recognition systems can be used to allow access to an ATM machine or a personal computer, to control the entry of people into restricted areas, and to recognize people in specific areas
(banks, stores), or in a specific database (police records). But in recent years face sketch recognition has become an active research area for engineers and scientist because an important application of face recognition is to assist law enforcement. Automatic retrieval of photos of suspects from the police mug shot database can help the police narrow down potential suspects quickly. However, in most cases, the photo image of a suspect is not available. The best substitute is often a sketch drawing based on the recollection of an eyewitness. Therefore, automatically searching through a photo database using a sketch drawing becomes important. It can not only help police locate a group of potential suspects, but also help the witness and the artist modify the sketch drawing of the suspect interactively based on similar photos retrieved [1], [2], [3], [4], [5], [6], [7]. However due to the great difference between sketches and photos and the unknown psychological mechanism of sketch generation, face sketch recognition is much harder than normal face recognition based on photo images. It is difficult to match photos and sketches in two different modalities
Related Work
For every unknown person, his/her face that draws our attention most. So face is the most important visual identity of a human being. For that reason face recognition has been an active research area over last 30 years. But in recent years sketch face recognition has been an active research area for engineers or scientists. Because in most cases the photo image of a suspect is not available in the police mug shot database. The best substitute is often a sketch drawing based on the recollection of an eyewitness. Therefore, automatically searching through a photo database using a sketch drawing becomes important. X.Wang and X.Tang [10] [11] proposed methods on sketch face recognition. But for the great difference between photos and sketches they first synthesize them and then applied most of the proposed face recognition approaches in a straightforward way.
Our Work
The human face has certain visual features that are common among everybody and some others that exhibit the unique characteristics of an individual. The system that we would like to develop has five components: 1) extract facial components from face photos and face sketch images 2) compute the local binary pattern for each facial component 3) extract curvelet features from face photo and face sketch 4) apply some classifier (like minimum distance classifier or ANN) in order to get best matching face photos corresponding to the probe face sketch 5) display the face photo which is best matching with input face sketch.
At present we have done some pre-processing work on face photos. Each of the captured 24-bits colour images have been converted into its 8-bit gray scale image. Gray scale image is then converted into binary image. The resultant image replaces all pixels in the gray scale image with luminance greater than mean intensity with the value 1 (white) and replaces all other pixels with the value 0 (black). In binary image, black pixels mean background and white pixels mean the face region.
Motivation
How do humans identify individuals with remarkable ease and accuracy? This question has haunted psychologists, neurologists, and recently engineers for a long time. The human face is different from any other natural or man-made objects, but has similar structural features across different races, sex and regions. The subtle variations in the face structure are captured by the human brain and help in discriminating one face from another. The human brain is able to filter out the common visual features of a face and retain only those suitable to exhibit the unique characteristics of an individual. Like face recognition, face-sketch recognition is also an important in today life because an important application of face recognition is to assist law enforcement. Automatic retrieval of photos of suspects from the police mug shot database can help the police narrow down potential suspects quickly. But in most cases, the photo image of a suspect is not available. The best substitute is often a sketch drawing based on the recollection of an eyewitness. Therefore, automatically searching through a photo database using a sketch drawing becomes important. But due to the great difference between sketches and photos and the unknown psychological mechanism of sketch generation, face-sketch recognition is much harder than normal face recognition based on photo images. It is difficult to match photos and sketches in two different modalities. Face-sketch recognition is not only help police locate a group of potential suspects, but also help the witness and the artist modify the sketch drawing of the suspect interactively based on similar photos retrieved.
Preprocessing
In the preprocessing step, photos are in colour, we first convert the RGB colour space to the gray colour space and then all the face photos and sketch photos are converted into binary image. Figure 3.1 shows (a) colour image, (b) corresponding gray image, © corresponding binary image. Figure 3.2 shows (1) a face sketch, (2) corresponding binary image. Preprocessing steps include four major steps they are: (a) Binarization (b) Extraction of Largest Component © Finding the Centroid (d) Cropping of the Face image in elliptical Shape
Extraction of Largest Component
The foreground of a binary image may contain more than one object or components. The large one represents the face region. The others are at the left hand bottom corner and a small dot on the top. Then the largest component has been extracted from binary image using “connected component labelling” algorithm [13]. This algorithm is based on either “4-conneted” neighbours or “8-connected” neighbours method [14]. In “4-connected” neighbours method, a pixel is considered as connected if it has neighbours on the same row or column. This is illustrated in Figure 3(a). Suppose the central pixel of a 3 × 3 mask “ f ” is f (x, y), then this method will consider the pixels f (x+1, y), f (x−1, y), f (x, y+1), and f (x, y−1) for checking the connectivity of f (x, y). In “8-connected”method besides the row and columns neighbours, the diagonal neighbours are also checked. That means “4-connected” pixels plus the diagonal pixels are called an “8-connected” neighbour which is illustrated in Figure 3(b). Thus, for a central f (x, y) of a 3 × 3 mask “f ” the “8-connected” neighbour methods will consider f (x−1, y−1), f (x−1, y), f (x−1, y+1), f (x, y−1), f (x, y +1), f (x+1, y −1), f (x+1, y), and f (x+1, y +1) for checking the connectivity of f (x, y).
Cropping of the Face Region in Elliptic Shape
Normally, human face is of ellipse shape. Then, from the above centroid coordinates, human face will be crop in elliptic shape using “Bresenham ellipse drawing” [16] algorithm. This algorithm takes the distance between the centroid and the right ear as the minor axis of the ellipse and distance between the centroid and the fore head as major axis of the ellipse.
DISCUSSION
Due to the great difference between sketches and photos and the unknown psychological mechanism of sketch generation, face sketch recognition is much harder than normal face recognition through face photo images. Therefore, in this project we will try to reduce the modality difference between face photo and face sketch by transforming images into feature domain. Two types of feature sets (spatial domain features and frequency domain features) will consider to represent each image (face photo and face sketch). Finally some classifier will use to recognize probe face sketch.