Well, when you see the image on your hard disk, it is compressed - for example as PNG or JPEG file. When you load the image in C#, it is decompressed, so the pixels can be rendered. So in reality, the image consists of more bytes than are stored on your hard disk.
The C# object contains the uncompressed data, so roughly (for RGBA images), the size in memory should be (4 * width * height) + further data
needed by .NET. The binary formatter does not save the image (as PNG or JPEG), it saves the object that represents the image.