Distance from a point to a line: Visual walkthrough

Step 1 — What is a point, what is a line, what are we even measuring?

WHAT. We have a single dot in the plane. We call it $P$ . Its address — how far right, how far up — is written $(x_{1}, y_{1})$ . The little $1$ is just a name-tag so we don't confuse our point with the wandering points that live on the line.

We also have a straight line $L$ . We write it as $a x + b y + c = 0.$

Read this out loud: "a point $(x, y)$ sits on the line exactly when $a x + b y + c$ comes out to zero." The letters $a, b, c$ are fixed numbers that describe the line — its tilt and its position — and $x, y$ are the free coordinates of whatever point we test.

$a$ and $b$ together set the direction the line points.
$c$ slides the line toward or away from the origin.

WHY this form and not $y = m x + k$ . The slope form $y = m x + k$ breaks for a vertical line (slope is infinite). The form $a x + b y + c = 0$ never breaks — a vertical line is just $b = 0$ . Because we want every line, we commit to this form.

PICTURE. Point $P$ off to one side, the line $L$ marching across, and the thing we want: the shortest hop from $P$ to $L$ , drawn in red.

Figure — Distance from a point to a line

Step 2 — Why the shortest path is the perpendicular one

WHAT. Of all the ways to walk from $P$ down to the line, one is shortest. Claim: it is the one that hits the line at a right angle (a perfect corner, $9 0^{\circ}$ ).

WHY. Pick any other landing point on the line. Now you have a triangle: the perpendicular drop is one leg, the slanted walk is the hypotenuse (the longest side of a right triangle). The hypotenuse is always longer than a leg. So every slanted path loses to the perpendicular one. This is why the whole problem is really about the perpendicular direction.

PICTURE. The perpendicular drop in red; two slanted alternatives in black, each visibly longer, each forming a right triangle with the red leg.

Step 3 — The arrow that points straight across: the normal vector

WHAT. A vector is just an arrow: it has a direction and a length. The special arrow $n = (a, b)$ — built from the same $a$ and $b$ in the line's equation — points perpendicular to the line. It is called the normal vector.

WHY is $(a, b)$ perpendicular to the line? Take two points on the line, $R$ and $S$ . Both satisfy the equation: $a x_{R} + b y_{R} + c = 0, a x_{S} + b y_{S} + c = 0.$ Subtract the two lines. The $c$ 's cancel and you get $a run along line (x_{R} - x_{S}) + b rise along line (y_{R} - y_{S}) = 0.$ That left side is exactly the dot product of $(a, b)$ with the arrow along the line. A dot product of zero means the two arrows meet at a right angle (see Projection of Vectors). So $(a, b) ⊥ line$ . That is the whole reason $a$ and $b$ show up in the answer: they are the perpendicular direction we need to measure along.

PICTURE. The line, and the red normal arrow $n = (a, b)$ shooting straight out of it at $9 0^{\circ}$ .

Step 4 — Stand on the line, look at $P$ : the connecting arrow

WHAT. Plant your feet at any point $Q (x_{0}, y_{0})$ that lies on the line. Since $Q$ is on $L$ : $a x_0 + b y_0 + c = 0 \quad\Longrightarrow\quad a x_0 + b y_0 = -c. \tag{$\star$}$ Keep equation $(⋆)$ in your pocket — we cash it in at Step 6.

Now draw the arrow from your feet $Q$ to the point $P$ : $QP = (x_{1} - x_{0}, y_{1} - y_{0}) .$ Each slot is "target minus start": how far right and how far up you'd travel from $Q$ to reach $P$ .

WHY. We don't yet know the true perpendicular foot. But we do know how to reach $P$ from some point on the line. In the next step we'll strip away the useless sideways part of this arrow and keep only the part pointing across — which is the distance.

PICTURE. $Q$ sitting on the line, the black arrow $QP$ slanting up to $P$ , and the red normal $n$ still poking out.

Step 5 — Casting a shadow: the projection onto the normal

WHAT. The arrow $QP$ points partly along the line and partly across it. The "across" part is our distance. To extract it, we project $QP$ onto the normal direction — we ask: how much of $QP$ lies in the direction of $n$ ?

The tool that answers "how much of one arrow lies along another" is the dot product divided by the length of the direction arrow: $d = QP \cdot \frac{n}{∣ n ∣} .$

WHY the dot product and not something else. The dot product $u \cdot v = ∣ u ∣∣ v ∣ cos θ$ measures overlap of direction. When we dot $QP$ with the unit normal $n /∣ n ∣$ (length 1), the answer is literally the length of $QP$ 's shadow on the normal line — precisely the perpendicular gap. The sideways part contributes zero because it is at $9 0^{\circ}$ to $n$ (its $cos 9 0^{\circ} = 0$ kills it). That is exactly what we want: the sideways travel should not count toward "distance to the line."

PICTURE. The black $QP$ , its shadow dropped onto the red normal direction, and that shadow's length labelled $d$ .

Step 6 — Plug in and watch $c$ appear

WHAT. Write out the dot product $QP \cdot n$ term by term: $QP \cdot n = right-part \times a a (x_{1} - x_{0}) + up-part \times b b (y_{1} - y_{0}) = (a x_{1} + b y_{1}) - (a x_{0} + b y_{0}) .$

Now cash in equation $(⋆)$ from Step 4, which said $a x_{0} + b y_{0} = - c$ : $= a x_{1} + b y_{1} - (- c) = a x_{1} + b y_{1} + c .$

WHY this is the punchline. The mystery point $Q$ has vanished. Its coordinates $x_{0}, y_{0}$ were eaten by the relation $(⋆)$ , leaving only $P$ 's coordinates and the line's constant $c$ . This is why we could stand on any point of the line — the answer never depended on which one.

Divide by $∣ n ∣ = a^{2} + b^{2}$ (Step 3) and take absolute value (Step 5):

Every symbol is now earned: $a, b$ = normal direction, $c$ = the shift that ate $Q$ , the bars = distances can't be negative, the root = the normal's length.

Step 7 — The degenerate cases (so you never get surprised)

WHAT / WHY / PICTURE, three quick edge cases the formula still handles:

$P$ is on the line. Then $a x_{1} + b y_{1} + c = 0$ , so $d = 0$ . The red segment shrinks to a dot. ✓
Vertical line, $b = 0$ (e.g. $x = 3$ , i.e. $x - 3 = 0$ , so $a = 1, b = 0, c = - 3$ ). The slope form would explode, but here $a^{2} + b^{2} = 1 = 1$ and $d = ∣ x_{1} - 3∣$ — the plain horizontal gap. ✓
Horizontal line, $a = 0$ ( $y = k$ , i.e. $y - k = 0$ ). Then $d = ∣ y_{1} - k ∣$ — the plain vertical gap. ✓
Two parallel lines $a x + b y + c_{1} = 0$ and $a x + b y + c_{2} = 0$ : pick any point on the first, drop it into the second, and everything but $∣ c_{1} - c_{2} ∣$ cancels, giving $d = \frac{∣ c _{1} - c _{2} ∣}{a ^{2} + b ^{2}}$ .

The one-picture summary

Everything in a single frame: point $P$ , line $L$ , the red perpendicular drop $d$ , the normal arrow $(a, b)$ , and the shadow that turns $QP$ into the answer.

Recall Feynman retelling — say it back in plain words

A line has a built-in arrow that points straight across it, called the normal, and its two numbers are exactly the $a$ and $b$ in the line's equation; the length of that arrow is $a^{2} + b^{2}$ . To find how far my point $P$ is from the line, I stand on the line anywhere, draw an arrow up to $P$ , and ask "how much of this arrow points across, in the normal's direction?" — that's a dot product. When I compute it, the spot where I chose to stand cancels out (because it's on the line, so $a x_{0} + b y_{0} = - c$ ), leaving just $a x_{1} + b y_{1} + c$ . I divide by the normal's length to turn "how much overlap" into a true length, and I slap absolute-value bars on because a distance can never be negative. Done: $d = ∣ a x_{1} + b y_{1} + c ∣/ a^{2} + b^{2}$ .

Question :: Where does the $a^{2} + b^{2}$ in the denominator come from?
Question :: Why does the arbitrary on-line point $Q$ disappear from the final formula?
Question :: Why the absolute value bars?

Question

It is the length of the normal vector

(a, b)

, needed to turn the dot-product overlap into a true perpendicular length.

Question

Because

Q

lies on the line,

a x_{0} + b y_{0} = - c

, and substituting that cancels

Q

's coordinates, leaving only

c

Question

The projection can be negative depending on which side of the line

P

is on, but distance is never negative.

Connections: Distance from a point to a line · Normal Vector to a Line · Projection of Vectors · Perpendicular Lines · Distance Formula · Equation of a Line · Linear Programming