The size of the photographs captured by a camera, quite often does not match the requirements. The photographs may have to be sent in small sizes by email or uploaded to social networking sites. There are websites that won’t even accept images above a certain size. On the other hand, the photographs may have to be enlarged and printed. The size of the photographs in most conditions also does not tally with the required print size. What happens in all these situations is that the various algorithms take up this task of changing the size of the photograph to provide acceptable results. These algorithms may be a part of the operating system, website designing software, printer’s software or even as coding at the level of websites. The problem? These acceptable results are most often not the best.
Present-day digital cameras have really high resolution sensors which capture large image sizes. With improvements in technology these are becoming better day by day. The real-world requirements of these megapixels may not be so high but having spare megapixels in hand does come in handy occasionally. The challenge comes when these photos are to be printed or even more so when these are to be used for displaying on computers. A high resolution big sized image is slow to upload and depending on the monitors, quite frequently down-sampled in real-time to show up on the monitor. The down-scaling that happens for displaying these big photographs is not really good and the photographs end up losing their impact. On the other hand, as I have mentioned earlier, some websites down-sample the image even before it is stored in their servers.
The ideal way out is to resample the images on your own and get the quality that you want. In some cases, the size itself may have to be increased. This is also called the interpolation of the images. Printers do it automatically when a small-sized photograph is to be enlarged. This is generally acceptable but is never an ideal solution!
(The image on the left is a 100% crop from the photograph at the end of this article, which was resampled for printing at a large size and the one on the right is the same image which was allowed to get resampled on a social-networking site while uploading. The difference is minimal but try to notice the loss of details, sharpness and subtle changes in tonal values in the image on the right)
Resizing or Resampling?
Some photo-editors like Photoshop give an option of resizing which is different from resampling. Gimp calls the control as ‘Print Size’. In resizing the number of pixels is kept the same but the print size is changed. So in the resizing option, if a photograph is to be printed to a large size, the number of pixel density (ppi) may decrease while keeping the same number of pixels in the overall image. The pixel size changes to bring about this change in size. This is not the best way to print photos. Similarly reducing the photograph size by resizing again leads to problems by creating a mismatch between how the printer sees the pixels and how the photo-editor wants them to be seen.
The better option is to resample the image. So, what exactly is resampling? This is a method where the numbers of pixels of a photograph are changed. It is, therefore, more effective in the problematic situations explained above. Resizing may save time and provide good enough results but the best option is usually to resample. As with all rules, there are exceptions but generally speaking try to resample rather than resize. This reminds me of a rhyme that I read somewhere – Good, Better, Best; never let it rest until your good is better than your best, & your best is better than the rest. So stop resizing and start resampling.
The steps to resampling an image are very simple
Choose the resolution. For best quality prints, set it to the highest possible value supported by the printer. Remember that a printer may require more than one dot to represent a pixel and so check your printer’s manual to understand this. For example, consider a photo-printer that has a specification of 600 dpi and about 2 dots per pixel. For such a printer using a 300 ppi image is the best option. Keep the values same for X and Y resolution if your favorite program has the option to set them individually. As far as possible, change the density by a factor so that it is easier for the program to calculate and the ‘original’ pixels from the photograph are used. Reducing the pixel density of a 300 ppi to 150 ppi will retain every second pixel. If the reduction was done from 300 ppi to 140 or 160 ppi, the calculation will be not so simple and the quality will deteriorate. Regardless of what popular sites say, setting the resolution initially makes resampling easier.
Set the image size by pixels if you plan to use it for websites or showing on electronic displays and set the value by Inches or Centimeters if you are planning to get it printed. Select only one dimension and let the photo-editing program choose the other one while keeping the aspect ratio intact. Resampled images with even a slight change in the ratio look very unnatural. Do not change the aspect ratio. This is done by using the ‘chain link’ icon. Once the ‘chain link’ is broken the X and Y values do not change in the fixed ratio whereas when the ‘chain link’ is intact the values change in such a manner that the aspect ratio is maintained.
Select the resampling method based on your requirements and intended use. More on resampling methods is yet to come.
Resampling the images always causes a drop in the overall quality. The resampled images, therefore, require sharpening and other adjustments later on, to make up for the loss in quality. Do this as the last step before saving your new resampled photograph.
(Settings I use in Gimp for downscaling a photograph for use on this site)
Resampling or Interpolation methods :
Nearest Neighbor—simple resampling which has the fastest processing time. This is ideally used for hard-edge images. Most of the popular websites that resample the images when they are uploaded, use this method.
Linear or Bilinear—algorithmic resampling is best used for shrinking images. I prefer this method for use on websites and for sharing images across social networks.
Cubic or Bicubic—algorithmic resampling can be used for enlarging images. Resampling is smoother than Bilinear but has a slower processing time. Though a good method, I rarely use this in my day to day work. Sometimes when I do, I use this method for website use only (see the image above). There are further subtypes to it depending on your program. Some Photoshop versions have smoother, sharper and automatic sub-types. If you are using this method to resample, use the automatic subtype without any changes. This should work for most photographs.
Sinc or Lanczos 3—complex algorithmic is the resampling method that gives the best results but with the longest processing time. This is my preferred method for resampling when my photographs are to be printed later on. In Affinity Photo, this is available as ‘separable’ and ‘non-separable’; the latter gives marginally better results. In Gimp, the Lanczos method is given without any further options. When in doubt, use this method for resampling.
(Upscaling in Gimp for printing the image to 12×18 inch size on a photo printer. The original image was about 11×16.5 inch in size)
Changing the canvas size instead of image size (cropping)
Crop tool can also be used to change the dimension of the image but as it is obvious this is done by discarding the unused areas of the photographs. Another reason why photographs sent for printing should also have the desired dimensions. If the image gets cropped in a printer, the cropping is almost centered which may not be the ideal or what the photographer intended. If you have to print an 8×12 inch photograph on an 8×10 inch paper, crop manually and then print.
The workflow that I use for resampling
- Open the raw file
- Do image adjustments in the same sequence as given here – Post Processing RAW. Exposure compensation, Shadow and Highlights detail recovery, Curves, Noise reduction, Burning / Dodging, Use of airbrush, stamping or cloning tool, etc, followed by any special filters if required.
- Resample the image to target resolution and size
- Sharpening at 20-50% depending on the image requirements (Sharpening)
- Export the image as jpg without compression if it is to be shared on social-networks or sent by emails, and with compression (about 80-90% of the quality) if it is to be used here at Maini’s Mind.
Some more points to remember –
- Do not use a higher resolution than required, especially while printing. This can be counterproductive and the prints may come out looking soft.
- Do not replace the original images with the re-sampled ones. Save them as a separate copy.
- Run the noise reduction at full resolution if you have to, before re-sampling!
- Remember to sharpen images after re-sampling, while viewing them at 100% magnification.
(Photograph used in this article – Nikon Df with Nikkor 105mm micro f/2.8 AI-s,
Exposure settings – f/8, 1/250 sec at ISO 100)
By the way, I would welcome comments on the images I use in this site (Maini’s Mind). What in your opinion is recommended out of these options? Feel free to send me your opinion using the form on the Contact page.
- The image sizes that I am using now with 80-90% of quality (some amount of compression) so that the site loads faster and eats less of your internet bandwidth.
- Same image sizes but with no compression. The image quality will improve slightly but at the cost of speed at which the site loads and also increased bandwidth consumption.
- Larger images with some more compression.
- Smaller images with no compression.
(Sorry! Being an ad-free site, I can’t afford to upload large images with no compression.)