A method by which a decoding device decodes an image includes the steps of: receiving a bitstream including residual information; deriving a residual sample for the current block on the basis of a quantized transform coefficient; and generating a reconstructed picture on the basis of the residual sample for the current block, where the residual information includes a context-based coded context syntax element, where the deriving the quantized transform coefficient decodes the context syntax element based on a context and based on a predetermined maximum value for the context syntax element, and where the maximum value is determined by a unit of a transform block.