Picture resolution at enlarged sizes

      A digital picture is composed of a number of tiny squares called pixels, which are at different shade of colors ( For a color image, there are 3 layers of pixels on top of each other, namely red, green and blue, and the number of shades per color is 256 if it is 8 bits colors . The color at each pixel is the combination of red, green, and blue at different shades ) These tiny squares form a picture like a jigsaw puzzle does. The resolution R of the picture is the number of pixels per unit length ( dpi, dots per Inches ) . Let W and H be the width and height of the picture. The number of pixels along the width is N = WxR and along the height is M = HxR.. The total number of pixels of the picture is P = NxM = WxH x R^2  ( For example, a 6 M pixels digital camera would give you  2000x3000 pixels pictures.

      It is important to know that for a picture with certain number of pixels, the resolution decreases at enlarged size. For example, if you have a 600x800 pixels digital picture, the resolution at 6"x8" size is 100 dpi. But if printed at 12"x16" size, the resolution will be 50 dpi ( remember R = N/W or M/H ). Therefore, to enlarge a picture,  the important information is the total number of pixels.  dpi is meaningless unless it is given at the print size. The higher the number of pixels, the sharper the image is. But please note what I meant was the number of pixels of the original images. You can easily increase the number of pixels through software interpolations. What it does is to break up the pixel squares into smaller ones based on the surrounding pixels information. It will reduce the pixelate appearance of the enlarged image. But it won't improve the sharpness and details of the image much. In fact, before printing, we always increase the number of pixels of the pictures if they are too low ,( and apply some sharpening to help the sharpness out a little ) and you don't have to do it at your end to save the uploading time.

     Enlarging vector based graphical elements ( Text fonts are one kind of them ) is another story. You can enlarge them at any size, without losing any sharpness and smoothness, as long as they were not rasterized ( turn into bitmap image ). They are stored by information ( coordinates, curvature ... ) of lines and curves. The software regenerate the objects when opening the file, at whatever resolutions they are being displayed and printed. Usually a graphic file contains both vector based elements and bitmap images. If you zoom up the graph, you will see the bitmap images become blur or pixelate, but the vector based elements remain smooth and sharp. There are many file types which can store vector based elements, including EPS, AI, PDF, CDR, PPT, DOC, PUB, WMF. PSD ....

     Bitmap images can be saved in file formats such as JPG, TIF, BMP ...  When saving in JPG format, the software offers different level of compression to reduce the file size. The compression is done by throwing away higher orders information in the picture and it is a lossy process. So always choose the high quality setting in the option windows ( Level 10 or higher in Photoshop would be sufficient ) to minimize the compression artifacts. ( Note: If the picture is already saved in highly compressed format, you can't reverse the loss by saving into a lower compression format ) For the best picture quality, saving in TIF with LZW compression is a good choice. It is a lossless compression, such as the compression in ZIP archives. The file size is larger than that saved in JPG in most of the case. But if the picture contains a large amount of uniform color areas, TIF is equal or better than JPG in term of file size reduction.

Film transfer : matching of frame rates

      Most of the home made films ( super8, 8mm, 16mm ) were recorded at 18 fps, while the NTSC video system has a frame rate of 30 fps or 60 fields per sec ( Each video frame is divided into two fields interlacing each other ). There are black-out periods between frames for the running films and the video. For free-run capturing process ( The process we are using ), it is important to synchronize these black-out periods. Otherwise, the transferred video will be blinking at low frequencies.  Each frame of the running films is divided into 3 frames by the black-out periods. The frame rate of the running film should be adjusted to 20 fps to sync with the video frames ( 20x3 = 60 ) . Therefore, the running speed of the films is slightly higher than 18 fps and it is neglegible for normal home movies. For PAL system video, the frame rate is 25 fps or 50 fields per sec. The running speed of the films is adjusted to 16.7 fps, and it is slightly lower than 18 fps. If the frame rate is critical and must be 18 fps, we have to apply an extra step using frame interpolation to convert the frame rate back to 18 fps. 

     Some service bureaus offer film transfers with frame to frame capturing process ( The cost is double compared to free-run transfer ). Unless they are using sophisticated video frame interpolation, they typically run the films at 24 fps and then use the so called 2:3 pull down ( 1st frame to 2 video fields, 2nd frame to 3 video fields and so on . This is the standard process used to convert the 24 fps telecine movies into NTSC video )  to convert to 30 fps video frame rate, and the film is running at a speed even higher than 20 fps.

     While running the film at a speed about 10% higher than it should be has a small effect on the video motion seen by human eyes, the audio tones on the sound track of the films is higher and is very noticeable. We apply software interpolation on the audio , preserving the tempo and lower the tone back to normal. The result is a correct audio tone with good synchronization between the audio and the video.