Pinkbike Photos - How to make them look amazing!

While Pinkbike is primarily a mountain biking website, it is also a site that involves a lot of photos taken by a great many photographers. With an average of over 4000 photos posted to the site every single day, Pinkbike is a busy place when it comes to pictures. While most of the images posted are shot by amateur photographers, there are also more than a few professional photographers on Pinkbike as well, and everyone regardless of their ability wants their photos to look as good as possible. So to that end I've put together a bit of info to help you make the photos you post on Pinkbike look the best.


In order to have your photos look as good as possible on the site please keep in mind the following rules.

1. Size your photos to either 1600 or 1024 pixels on the long side. This is very important. 800 pixels also works, but in this day of larger screens most people are disappointed if a photo is only available as an 800 pixel 'thumbnail'

2. Make sure your photo looks great at whatever size you upload it (1600 or 1024) If the original is oversharpened, Pinkbike won't make it look any better. And the same goes for images that are soft, make sure the original looks good if you want it to look good on Pinkbike.

3. Embed the sRGB color profile in the image. If you want the colors to look accurate on the web make sure to embed the sRGB color profile into the image, Firefox and Safari recognize color profiles and use them, rendering proper colors.

4. Use Firefox or Safari to view images on the web, IE7 if you need to use a Windows product. Google Chrome does not have any color management as of this post, and will not give you accurate color no matter what you do to your photos.

5. Also please be sure to include your name and or copyright in the Metadata of the photo if possible.


Image Quality and Resizing

The past year I've spent a bit of time working on the image resampling and sharpening algorithm with Pinkbike's resident tech guru, Radek.
During the time that I've been working for Pinkbike there have been a few comments about images becoming over sharpened when uploaded to the site. And also the opposite, some people have said their images were made softer when they uploaded them. That's two completely opposite things, so Radek and I sat down to figure it out. Radek spent a lot of time configuring the original algorithm, and he did a lot of research. He even discovered a few things that I as a photographer really knew very little about. As a result of that the original resampling/sharpening code on the site is quite advanced, possibly even ahead of photo sharing sites such as Flickr in some ways. When I looked into what was happening I took a bunch of different examples I had seen or been sent, and looked at the original images as well as the resampled ones on Pinkbike. Then I uploaded different images and compared how they looked at Pinkbike's various photo sizes to how I wanted them to look. The results were fairly surprising. To understand them you need to know a little about photo resampling to start with. Here are the basics...

To start with some of you are probably wondering why I'm saying resampling instead of resizing. Resizing an image refers to changing the size it will print, without changing the actual pixels. Pinkbike doesn't ever make photos bigger, only smaller. When you make an image for the web smaller, pixels are removed from it to make it physically smaller, and this is normally referred to as downsampling. Which pixels are removed depends on which interpolation method you use to resample the image. Regardless of which method you use for downsampling an image, when you remove pixels edge detail is lost, and it ends up making the image appear softer. Photoshop gives a few different options depending on which version you're using, most people never even notice them and just go with the default. Here is the description that Adobe gives for the relevant downsampling options available in PS CS5:

Bicubic
A slower but more precise method based on an examination of the values of surrounding pixels. Using more complex calculations, Bicubic produces smoother tonal gradations than Nearest Neighbor or Bilinear.

Bicubic Sharper
A good method for reducing the size of an image based on Bicubic interpolation with enhanced sharpening. This method maintains the detail in a resampled image. If Bicubic Sharper oversharpens some areas of an image, try using Bicubic.

There are other options such as Lanczos or Sinc that work better (on Pinkbike we use Lanczos), but for the purpose of downsampling in Photoshop these are usually the only two that we use. Bicubic produces the nicest gradations but tends to make images look a little softer while Bicubic Sharper has built in sharpening to improve this. The problem with Bicubic Sharper is that while it works well for some images it can really oversharpen certain details and textures. The best way to downsample for our purpose is to use Bicubic, and then manually sharpen after as needed. This usually works really well, however it is image and size specific. Some images need more sharpening than others, and images need different amounts of sharpening depending on how much they've been downsampled.

On a website there is no way to control the amount of sharpening that is applied to a specific image, so we need to assume all images that are uploaded are sharpened properly to begin with. That gives us a neutral starting point with which to figure out size specific sharpening. When I tweaked the sharpening algorithm I uploaded several different images that were 1600 pixels on the long side (Pinkbike's recommended size) and adjusted the sharpening on each of the set image sizes until they all looked good. If you're having problems with the way the website sharpens your images make sure you haven't oversharpened them to start with, and be sure to upload them at 1600 pixels on the long side if possible. The algorithm that we use to downsample and create the various sizes used on the site (1024, 800, 500, and 250 pixels) is designed to work with a starting size of 1600. It applies the same sharpening to an image downsampled to 1024 pixels regardless whether the starting size was 1600 or 1100. This is the same for all image sizes. So a starting image of 1100 pixels will produce a 1024 pixel image with more sharpening than a starting image of 1600 pixels will. An image that is uploaded at 1024 or 800 pixels will look great at that size as no processing, downsampling, or sharpening is done to it, however it may be slightly oversharpened when it's downsampled to the smaller sizes.

Gamma Correction

There's more to downsampling than just simply removing pixels and then sharpening to bring back sharpness. If you take an 8 bit image in Photoshop and just downsample it 50% you physically lose a lot of the fine detail. This has to do with the way that most image resampling software interprets the brightness scale as being linear, when in fact it is not. A great example would be this image of the Dalai Lama that was modified by Eric Brasseur. He added a gray line screen to it that changes the way it looks when it's resampled. Try downloading and downsampling this image in photoshop and see what happens. Or if you can't be bothered to do it yourself take my word for it and look at the examples below. You can see the original image on the left, beside it is the image as it appears on Pinkbike at 150 pixels, and then next to that is the image that I downsampled to 150 pixels in Photoshop. Interesting isn't it...