Megapixels and more. The basics of image processing in digital cameras. Construction of a digital camera

Transcription

1 Megapixels and more The basics of image processing in digital cameras Photography is a technique of preserving pictures with the help of light. The first durable photograph was made by Nicephor Niepce in 1826 (179 years ago). The picture was of not so good quality, and it was obtained on a metal unpolished plate. The light-sensitive material consisted of petroleum components. In 1839 Louis Jacques Daguerre invented the DAGUERREOTYPE, which enabled to receive silver pictures on copper plates; then William Fox Talbot invented a negative, which is the basis for traditional photographic techniques. The first color picture was made by James Clerk Maxwell in The first color film Autochrome, being based on the potatoes starch, was introduced in 1907, and the first threelayered color film Kodachrome was produced in Until the 1990s, the traditional techniques were only being improved by introducing new technical solutions, facilitating the performance of correct pictures and increasing the quality of imaging. The development of electronics in the 1980s and mass application of integrated circuit technology enabled to construct a cheap picture converter, which converted the photon flux into an electric signal. It followed that the first digital cameras with digital memory cards instead of a typical film entered the market and the revolution in technology of preserving pictures began. Classification of digital cameras: Compact cameras with non-interchangeable fixed focus, Compact cameras with non-interchangeable zoom lenses, Hybrid cameras with non-interchangeable zoom lenses, Reflex digital cameras with interchangeable lenses (DSLR, Digital Single Lens Reflex Camera). Construction of a digital camera Sensor There are two types of optical converters in digital cameras (sensors or matrixes): CMOS less popular, CCD more popular. The invention of CCD goes back to the beginning of the 1970s.

2 The CCD sensor can be simply defined as one large siliceous diode. The base is a semiconductor type p, the upper layer is type n, and there is a potential barrier between them. On such a wafer a thin layer of insulator is applied. The light which falls on the silicon causes the knocking-out of single electrons from their orbits and generating the pairs of a holeelectron in this way. The electric field of potential barrier divides these pairs and relocates the electrons towards the area - type n, and the holes towards the base type p. The holes float away into the base and the electrons are trapped in the semiconductor - type n. The electrons are spread over the surface of insulator. The longitudinal and narrow strips of metal which run straight along the width of sensor are charged positive by control systems, therefore the electrons will gather underneath. In order to immobilize the electrons, narrow strips of admixture are inserted additionally in the semiconductor, which divide the surface of sensor into separate lines. The number of electrons gathered under electrodes is proportional to the local lighting of its area. In order to insert these electric charges, two additional electrodes located between the main electrodes are needed. They are not usually charged, and then they do not have an effect on the distribution of electric charges. While the matrix is being read, they are being charge one by one causing relocation of electric charges from the center to the edges of matrix. The whole line of matrix is transported in one move. At the edge of matrix the charges are moved similarly from one line on the right to the other one on the left (or the opposite). An electrical signal which is received from one matrix cell is intensified and converted into a digital value. In the beginning the CCD sensors dominated the market regarding a better quality of picture. A disadvantage of CCD is its construction: in order to measure the charge, it needs to be moved out of the area of matrix, which is connected with the loss of electrons and related noise. Due to the process of production, the electronics controlling the matrix in the same integrated circuit cannot be integrated; therefore, additional elements controlling the matrix are needed. The CMOS-APS (Active Pixel Structure) sensor is constructed of the base type p, in the middle of each pixel a chip of semiconductor type n is put in creating a photodiode. At the edge of a pixel three MOS transistors are located. At the request T1 connects the photodiode to + supply resetting a pixel, T2 changes the accumulated charge into the proportional electric current, at the request T3 connects the outlet of the transistor T2 to the matrix outlet allowing to choose which line is to be read in a given moment. 2

3 The T3 transistor gates of every pixel are connected by lines, so the whole line always becomes activated. The same applies to the T1 controlling gates the whole line is erased at once, usually right after it has been read. A given line can be read many times, the reading does not remove the electric charge. A disadvantage of CCD sensor is the occupied silicon area. A typical 6 Mpix sensor requires 6 million photodiodes and 18 million transistors (more than many processors have). A significant advantage is the location of matrix and controlling systems in one integrated circuit, which reduces production costs and the price of the final product. The human eye has three sensors reacting to colors of objects: sensitive to red, green and blue colors and one reacting to the intensity of light (rod cell of retina). Similarly, in a camera there are filters separating these three colors from the light reaching the sensor Since every cell of the optical sensor reacts only to the intensity of light falling on it, not to its color, it is necessary to de-filter an adequate component of light color which falls onto the sensor. The Bayer filter is the one which is the most often used, in which lines of green and red filters alternates with lines consisting of green and blue filters. As it can be seen, there are twice as much cells reacting to the green component, then those which are sensitive to red and blue. By way of example, there are about 2 million cells sensitive to green and 1 million cells sensitive to red, and 1 million cells which are sensitive to blue in the matrix of 4 Mpix. It results from the fact that the human eye is more sensitive to green and yellow components, thanks to which we see more details in these two colors. Every optical sensor cell supplies information about the intensity of light of one color component, while the others must be calculated on the basis of information reaching the adjacent cells, which react to the other color components. This is an interpolation process, which is also called demosaicing. An interesting solution is offered by the matrix of Foveon type, in which semipermeable layers were applied, which are sensitive to blue, green, and red colors. Thanks to that every cell of sensor directly supplies information about the three colors. One of the most popular mistakes which are being made while choosing a camera, is the carelessness about the influence of size of the optical sensor (matrix). A typical sensor measures from about 4 x 3 mm to about 9 x 7 mm (it is 2 x 3 mm in a mobile telephone).therefore, there are 4 million cells located on the area from 12 mm 2 to about 63 mm 2. It is easy to calculate that one cell takes up an area from 3 m 2 to 16 m 2. The smaller the area of the cell, the fewer number of photos will reach it; thus, the smaller the useful electric signal will be. While the size of cell becomes smaller, the signal will be 3

4 become close to the noise level of matrix and the analog part of electric circuit of camera. Therefore, some cameras are characterized, among other things, by a higher level of noise then others. An observable tendency is increasing the resolution of matrix (the megapixels) without increasing its physical size. It can be easily guessed how the quality of taken pictures is affected. Producers aim at lowering the level of noise by applying stronger noise reduction algorithms; however, it also causes the loss of details of recorded pictures, so the increasing resolution of matrix cannot be fully used. Only the reflex cameras offer matrixes which measure from 24mm x 18mm to 24mm x 36mm (just as a typical 35mm film). There are also bigger matrixes in digital reflex cameras, but their prices go up along with the increase of size of the converter. The most of reflex cameras for advanced amateurs and semi-professionals have an APS-C matrix measuring around 22mm x 18mm. While comparing the dimensions of matrixes, one can notice that the size of a singe cell in a reflex camera is many times bigger than in other cameras, thus the noise is less visible. Optics Every optical system has two adjusting parameters: focal length and aperture. If lenses are not correctly focused on a photographed object, every point of the object creates a picture in a form of a round speck instead of the point and the whole picture is blurred. In a camera the focusing is made by moving the lenses to or away from the sensor. There are the following types of lenses: fixed focus, zoom lenses. The aperture adjusts the amount of light that reaches the optical converter, which means that it also influences the brightness of picture created on the converter. The parameters of the optical system mutually affect one another in different ways. For example the amount of light and sensitivity of optical sensor affect the sharpness of photographed objects. Advanced photographers know that every lens captures more clearly, when it has a small aperture. The other feature of aperture size is the so-called depth of field. A lot of light in the surrounding or a good sensitivity of optical sensor allow for big values of aperture, which additionally besides, increasing the sharpness of photographed objects increases the distance, within which the objects are sharp and clear. However, opening the aperture leads to decreasing of the depth of focus. An experienced photographer who plans to take a picture, considers not only the contents of a frame, i.e. a reasonable placement of 4

5 objects in two dimensions, but he also builds the so-called picture plans by applying skillfully the depth of focus in order to bring out the two main themes and a blurred background. The adjusting aperture is essential both in the human eye and camera, because the range of changes of brightness in the surrounding is significantly bigger than the range which is directly converted by the optical converter. The range of changes of brightness between the lighting of object on a bright sunny day and the lighting of the same object by the moonlight comes to one million. Having considered the ability to reflect the light by an object within 1% to 99%, we will obtain a difference in the intensity of light equaling one to a hundred million. The dynamic range of camera s sensor, which is defined as the value ratio of the biggest converted signal to the level of matrix noises, usually equals 300 to In other words, if the biggest converted signal has the amplitude of 1 V, then the level of matrix noises will reach 1mV. A camera has an aperture which can attenuate the light reaching the sensor by a coefficient 300. It follows that the range of dynamics typical for a camera comes to a few hundred thousand. Signal processor and software The matrix of camera is an analog device. The first stage of signal processing consists of its intensification and converting it into a digital signal by the analog-digital converter. Next, the received beam of bits can be processed by the signal processor. The first stage of processing is the above-mentioned demosaicing. Then, there is an initial sharpening of the picture (producers of cameras reluctantly admit that), which is followed by the process of recoding of raw image (RAW), or its compression to JPEG format (the most often), i.e. picture is divided into groups of 8 x 8 pixels and the image is discreetly transformed DCT (Discrete Cosine Transform), after that depending on an accepted degree of compression a proper quantization table is applied, and in the end there is a suppression of repeating zeros with the help of RLE algorithm and the Huffmann coding. Afterwards, the image is stored in the camera. The camera software also offers other additional functions, such as service of camera manipulator, automatic picture sharpening and selecting the exposure time, LCD picture display, and transfer of pictures into a computer. 5